Aminet 1 (Walnut Creek)/Aminet - June 1993 [Walnut Creek].iso/aminet/misc/sci/dproc126.lha/docs/DataProc.doc

Text File | 1993-02-28 | 114KB | 2,835 lines

1 dataproc
?
 DataProc is a command-driven data processing program, which works much
 like a tabular calculation program. DataProc reads in a command file
 holding commands and data and can produce a number of different output
 files for further processing with other software. DataProc is especially
 designed to work with GNUPLOT and LaTeX. Both of the latter are available
 as public domain software on almost any computer system.

 Beyond the standard mathematics to apply on data several statistic specials
 like regression and averaging are implemented.

 You should pass a filename to DataProc. This file wil be read and executed
 by DataProc. The default extension for this file is '.tst'. So DataProc
 tries to open <filename.tst> if <filename> is not available. E.g.:

            DataProc project.tst
            DataProc project

 are equivalent, unless a file named 'project' exists, which will then be
 read by the second line. If the file 'project.tst' does not exist an error
 will be issued.

 If you don't pass a commandfilename, DataProc will enter the interactive
 mode. Anything you enter behind the prompt will be executed just like
 a commandfile line. As comments don't seem to make sense in interactive
 mode, any comments will be interpreted as include file names. See
 `commands include` for more information. An example session might be:

            TeX:physics> DataProc
            DataProc 1.221 (24.1.93)
            © 1992 by MGR, Asgard
            This Amiga copy is registered to Lars Hanke #: 000012

            1 > *project1
          [messages from execution of project1.tst]
            2 > *project2
          [messages from execution of project2.tst]
            3 > #quit

            TeX:physics>

2 politics
?poltics

 Some words about software in general. Never use original files to run
 software. Always create copies and run them, because you will never
 know what might happen to the software.

 If you have problems getting started with the evaluation copy, contact
 me immediately, no matter if you want this program or not. I cannot
 test the software on a lot of systems with different configurations
 and it may do really weired things on other systems. If I do not come
 to know it, this won't change ever and at most leads to a bad reputation
 of the program.

3 distribution
?distribution

 In a nutshell, DataProc may be distributed freely as long as
 the following restrictions are met:

 The distributor may only charge a fee up to the costs of
 obtaining a public domain diskette from Fred Fish. The
 distributor agrees to cease distributing the programs and data
 involved if requested to do so by the author. The distributor
 may only distribute an unmodified copy of the original program,
 with all the supplied documentation and copyright notices
 intact.

 For more details, see the section called `License`.

3 support
?support

 If you have any suggestions, bug reports etc., feel free to
 contact the author of DataProc at the address given in the
 section "Registration". Additionally, you can reach me at:

          Lars Hanke
          FIDO:   2:245/5802.22
          UUCP:   [e-mail]

3 license
?license

 1. This license applies to the product called "DataProc", a
 program available for the Amiga computer and MS-DOS (tm) machines,
 published by Lars Hanke under the concepts of ShareWare, and
 the accompanying documentation, example files and anything else
 that comes with the original distribution. The terms "Programs"
 and "DataProc" below, refer to this product. The license is
 addressed as "you".

 2. You may copy and distribute verbatim copies of the
 programs' executable code and documentation as you receive it,
 in any medium, provided that you conspicuously and appropriately
 publish only the original, unmodified programs, with all
 copyright notices and disclaimers of warranty intact and
 including all the accompanying documentation, example files and
 anything else that came with the original.

 3. You may not copy and/or distribute these programs without
 the accompanying documentation and other additional files that
 came with the original. You may not copy and/or distribute
 modified versions of these programs.

 4. You may not copy, modify, sublicense, distribute or
 transfer the programs except as expressly provided under this
 license. Any attempt otherwise to copy, modify, sublicense,
 distribute or transfer the programs is void, and will
 automatically terminate your rights to use the programs under
 this license.

 5. By copying, distributing and/or using the programs you
 indicate your acceptance of this license to do so, and all its
 terms and conditions.

 6. Each time you redistribute the programs, the recipient
 automatically receives a license from the original licensor to
 copy, distribute and/or use the programs subject to these terms
 and conditions.  You may not impose any further restrictions on
 the recipients' exercise of the rights granted herein.

 7. You may not disassemble, decompile, re-source or otherwise
 reverse engineer the programs.

 8. You may use the programs for a period of up to 40 days for
 evaluation.  After that, you have to register.

 9. If you wish to incorporate parts of the programs into other
 programs, write to the author to ask for permission.

 10. You agree to cease distributing the programs and data
 involved if requested to do so by the author.

 11. You may charge a fee to recover distribution costs. The
 fee for diskette distribution may not be more than the cost to
 obtain a public domain diskette from Fred Fish.

3 warranty
?warranty

 THERE IS NO WARRANTY FOR THE PROGRAMS, TO THE EXTENT PERMITTED
 BY APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE
 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAMS "AS
 IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED,
 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE
 ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAMS IS
 WITH YOU.  SHOULD THE PROGRAMS PROVE DEFECTIVE, YOU ASSUME THE
 COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.

 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY
 REDISTRIBUTE THE PROGRAMS AS PERMITTED ABOVE, BE LIABLE TO YOU
 FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
 THE PROGRAMS (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA
 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
 PARTIES OR A FAILURE OF THE PROGRAMS TO OPERATE WITH ANY OTHER
 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED
 OF THE POSSIBILITY OF SUCH DAMAGES.

3 disclaimer
?disclaimer

 No warranty, either express or implied, is made with respect
 to the fitness or merchantability of DataProc.

 Lars Hanke (referred to as "the author"), reserves the right to not
 develop any future versions of DataProc.

 The author will try to make a good faith attempt at
 correcting any problems if any are discovered, but is in no way
 required, nor bound to correct them.

 The author neither assumes nor accepts any responsibility for
 the use or misuse of these programs. He also will not be held
 liable for damages or any compensation beyond the original
 registration fee due to loss of profit or any other damages
 arising out of the use, or inability to use these programs.

 Lars Hanke will not be liable for any damage arising from the failure
 of these programs to perform as described, or any destruction of other
 programs or data residing on a system attempting to run the programs.
 This does not depend on my knowledge of damaging errors in the programs.

 I may change prices and distribution conditions at will, with every new
 revision of the programs. I cannot be forced to distribute old revisions.
 Registrations not meeting the current conditions will lead me to inform
 you of the changes. You may then cancel your registration or meet the
 new conditions. Note that you have to express any interest in special
 distribution conditions explicitly, or I will assume that you agree with
 the current conditions. Lars Hanke is not liable for any costs of monetary
 transfers and other expenses concerning a cancelled registration.

3 registration
?registration
?shareware
 As already mentioned, this program is Shareware. Shareware is neither
 Public Domain, nor Freeware, nor Giftware, nor Copyleft. If you decide
 to use this program, you must pay for it! If you don't pay, but use, this
 is software piracy, with all its legal consequences.

 Some reasons to register: First of all, to stay legal, but as nobody cares
 nowadays, there are some other good reasons. An important reason is the
 support of the shareware idea. You don't buy shareware blindly, but you
 may test it thoroughly before. But if nobody pays, it is only a logical
 consequence that no-one will write shareware anymore and finally you have
 no chance, but buying software blindly. Finally the support of Shareware
 programmers will encourage them to care for their programs and you are
 more likely to get the same support as with commercial software.

 Unfortunately, experience tells us that those reasons aren't too convincing
 for a lot of users. Well, here are some reasons to register DataProc.
 First of all, the banner message will change from "unregistered" to your
 name and registration number. Then you will be able to install DataProc
 anywhere in your path, the evaluation copy may only be started form its
 own location directory. Finally those nerve-racking Shareware posters
 with its requests for keyboard input during the execution of your commandfile
 will disappear. The evaluation copy of DataProc is fully operative, but
 has limitations to the possibilities of automatic execution e.g. during
 batch jobs and to a proper installation in your system, e.g. paths or
 packing of the program. Due to a loss of dumb checks whether the copy may
 be used or not, the registered software improves a great deal in performance,
 especially the reloading of the own code in bytes will drop.

 Remember, a software worth using is a software worth buying. And don't tell
 me you can't afford DM 25.--, you could still sell your computer! This
 price is for registration only. The keyfile will be sent via UseNet. If
 you don't want your keyfile via UseNet or would like an update included
 add DM 2.-- for European Union countries or DM 5.-- for countries outside
 the EU, and I will send a disk including your keyfile and the most recent
 DataProc version. Note: If the service "keyfile via email" becomes too
 excessive, I may quit that service at will, either mine or the will of
 my mailservers.

 Send your registration (including: Name and address, DataProc version,
 source you got DataProc from, your system configuration and your EMAIL
 address) to:

         MGR, Asgard
         Lars Hanke
         Schlachthofstr. 67
         D-4690 Herne 2

         FIDO: 2:245/5802.22
         UUCP: [e-mail]

 This is also the adress to send suggestions and bug reports to. For urgent
 bug reports try:

               [e-mail]

 If you intend to install this programm on a multi user system, or are
 interested in ports to other machines or want to customize it, send me a
 note. I will gladly cooperate.

3 credits
?credits
 Thanks to the developers of GNU-Plot and TeX. The existance of those
 products was the main influence to start this project. Also the basic
 layout of the user interface was drawn from there. This does of course
 in no way apply to any code used for DataProc.

 Thanks to TrapLand, showing me that SHAREWARE might work even in Europe.
 I hope you don't mind that I took your legal stuff as the basis of the
 DataProc license.

 Thanks to Mr. Teulings, Mr. Preuß for beta testing and Mr. Kulka for
 alpha testing. Thanks as well to Mr. Böllmann and Mr. Beisemann for
 lending me their machines for the MS-DOS compilation.

 Thanks to Mr. Teulings for "WRITE" coping with those goddamned space
 syntax in the .doc file.

 Thanks to all those helpful guys round FIDO and UseNet, helping me
 with coding and porting problems.

 Thanks to Commodore for producing the Amiga and to SAS for their
 prompt help and support of their compiler.

 Finally thanks to me for not needing any sleep at all and to my dog
 for generously freeing me of that task, sleeping for both of us.

 Thanks to Odin, the Aesir and the Runes for my name and logo.

2 basics
3 syntax
?commandfile
?commandline
?comment
?grouping
?seperator
?token
?syntax
 There are some basic rules about the structure of command files. In general
 empty lines and the number of spaces are ignored. The same is with leading
 and trailing spaces, tabs or whatsoever. DataProc does not distinguish
 between different spacing characters.

 The characters '*', ';' and '%' mark a commentline, which is ignored. Inline
 comments are not yet supported. Note that in interactive mode the line
 following the comment sign will be passed as arguments to the include
 command. See `commands include` for more information.

 The characters '!' and '$' will post whatever they are followed in that line
 on screen during the execution of the commandfile. They should be used to
 supervise proper operation of a commandfile.

 A commandline is made up by ASCII characters and ends with the EOL character
 (usu. $0a). By now you cannot mark a line as a follower.

 Commandlines are made up by tokens. Each token is seperated by either a space
 or ':', '/' or '&'. While '//' declares an empty token two spaces do not.
 (See above). Since Version 1.108 system tokens aren't case sensitive anymore.
 Strings you define yourself of course won't be translated to a special case.

 Tokens may hold seperators, if they are enclosed in grouping characters.
 These grouping character pairs are: () or [] or '' or "". The grouping
 pairs may not nest. If an opener (bra) is encountered, the input line will
 be copied to the token until the corresponding close character (ket) is
 found.

 Commandlines hold either a command or data. The line is interpreted as a
 command, if it begins with a sharp sign '#'. In any other case DataProc
 expects data to be there.

3 organization
?table
?column
?internals
 Internally DataProc works much like a spread sheet program. It inits a table
 of a user defined length and can add an arbitrary number of columns to it.
 Commands usually affect a whole column. As you might guess the column is the
 basic unit in DataProc. That is the major difference to a tabular calculator
 where the cell serves as its unit.

 Each column has one physical and two logical numbers. The physical is the
 basic reference. It is simply the number of the column as it was declared
 by the `columns` command.

 Beyond that each column may be assigned a specific read column and a write
 column, meaning the order data is read into them or written from them. For
 more information see `using` and `ocol`.

3 mathematics
?function
?mathematic
 DataProc supplies a scientific calculator. You may define your own constants,
 variables and functions. Read the following for its syntax:

   ATOM        - Atoms are plain Values like 7.65e-12 and constants like PH
                 Functions are treated as atoms as well, but the arguments
                 are itself Expressions

   OPERATOR    - The currently supported operators are + - * / ^ with the
                 commonly known meaning. Missing operators like in 3sin(x)
                 or 3 4 are treated as '*' so that the former expressions
                 would be expanded to 3*sin(x) and 3*4. Operators obey the
                 common priority rules.

   BRACES      - grouping characters. You may set Expressions into pairs of
                 corresponding braces. All types are treated equally, but
                 must of course match. The brace pairs are: [] {} ()
                 Thus (3*[4+5]-2)/6 would work but (3*[4+5)-2]/6 would not!

   EXPRESSION  - any valid combination of Atoms, Operators and Braces.
                 ATOM OPERATOR ATOM
                 EXPRESSION OPERATOR EXPRESSION
                 ATOM ATOM                      => OPERATOR = '*'
                 EXPRESSION EXPRESSION          (dto.)
                 ([{ EXPRESSION }])             => treated as ATOM in this
                                                   level
   NAME        - Names start with an alphabetic character and may be followed
                 by any combination of alphanumeric characters. The total
                 length must not exceed 29 characters. Names are not case
                 sensitive.

   VARIABLE    - NAME followed by OPERATOR or end of line

   FUNCTION    - NAME followed by an opening brace. In defines this must be
                 an opening parenthesis '('. Note that a defined VARIABLE
                 in the combination VARIABLE(... will be correctly interpreted
                 as VARIABLE*(... and a defined FUNCTION in the combination
                 FUNCTION EXPRESSION will lead to the message --> Argument
                 expected. Note also that spaces are ignored, except for
                 separation numbers, so that 3 4 is interpreted as 3*4 and
                 34 is interpreted as 34.

   DEFINE      - ":=" defines or redefines a variable or function. When
                 defining a function the next character must be '(' (spaces
                 are ignored)
                 VARIABLE:=EXPRESSION                 => define a symbolic
                                                         variable
                 FUNCTION(VARIABLE,...):=EXPRESSION   => define a new function

                 Note that the template arguments are not stored to the
                 variable list. If identical entries exist in the variable
                 list they will not be recognized -- the template has higher
                 priority. Note as well that global variables are evaluated
                 immediately, so later changes won't affect the function.

   COMMENTS    - The line is terminated by any nonprintable character or a
                 semicolon <;>. The rest can be used for comments.

   RESERVED    - In future versions the following symbols may come to a
                 meaning, so do not put your own stuff on them! The charcters
                 are: ! _ _i _[]


 `SYMBOLS`

 The following symbols are predefined:

   PI          - you now that number (3.14...), don't you?
   ME          - the Euler number                  (2.718...)
   RT2         - the square root of 2              (1.41...)
   PH          - Planck's quantum                  (6.626...e-34)
   PHB         - Planck's quantum divided by 2*PI
   PE0         - electrical field constant SI      (8.854...e-12)
   PU0         - magnetical filed constant SI      (1.256...e-6)
   PC          - light speed in m/s                (2.997...e8)
   PK          - Boltmann constant                 (1.380...e-23)
   PR          - general gas constant              (8.31441)
   PSIGMA      - Stefan-Boltzmann constant         (5.670...e-8)
   PE          - charge of electron in C           (1.602...e-19)
   PME         - mass of electron in kg            (9.109...e-31)
   PBM         - Bohr's magneton in Am^2           (9.274...e-24)
   PMP         - mass of proton
   PMN         - mass of neutron
   PF          - gravity constant                  (6.672...e-11)
   PG          - standard earth acceleration       (-9.80665)
   PKEL0       - absulute zero temperature         (-273.15)
   PPN         - normal pressure in Pa             (101325)
   PVM0        - ideal gas molaric volume in l     (22.41...)
   PNA         - particles per mol                 (6.022...e23)
   PFARAD      - Faraday constant                  (9.648...e4)
   PRH         - Rydberg constant                  (1.097...e7)
   PLC         - Compton wavelength                (2.426...e-12)
   PSOLAR      - solar constant                    (1.368e6)

   N2KP        - newton to kilopond                (0.102)
   MS2KMH      - m/s to km/h                       (3.6)
   J2MEV       - Joule to MeV                      (6.241e12)
   TORR2P      - Torr to Pascal                    (133)
   U2KG        - u (nuclear mass unit) to kg       (1.660...e-27)
   KG2MEV      - kg to equivalent energy in MeV    (5.609...e29)
   U2MEV       - u to MeV                          (931.433...)
   LY2M        - light years to meters             (9.465...e15)
   AE2M        - astronomic units to meters        (1.495...e11)
   PARSEC      - Parsec to meters                  (3.085...e16)
   IN2CM       - inches to centimeters             (2.54)
   FT2CM       - feet to centimeters               (30.48)
   YD2CM       - yards to centimeters              (91.44)
   MILE2KM     - miles to kilometers               (1.609)
   PS2KW       - german horsepowers to kilowatts   (0.736)
   CAL2J       - calories to Joule                 (4.1868)

   EXA         - scaling syllable                  (1e18)
   PETA        -                                   (1e15)
   TERA        -                                   (1e12)
   GIGA        -                                   (1e9)
   MEGA        -                                   (1e6)
   KILO        -                                   (1e3)
   HEKTO       -                                   (1e2)
   DECA        -                                   (10)
   DECI        -                                   (0.1)
   CENTI       -                                   (1e-2)
   MILLI       -                                   (1e-3)
   MICRO       -                                   (1e-6)
   NANO        -                                   (1e-9)
   PICO        -                                   (1e-12)
   FEMTO       -                                   (1e-15)
   ATTO        -                                   (1e-18)

 Arguments of trigonometric functions are expected in radian!

   SIN(x)      - sine function                     sin(x)
   COS(x)      - cosine function                   cos(x)
   TAN(x)      - tangent function                  tan(x)
   COT(x)      - cotangent function                cot(x)
   SEC(x)      - secant function (1/sin)           sec(x)

   ASIN(x)     - arcus sine function               arc sin(x)
   ACOS(x)     - arcus cosine function             arc cos(x)
   ATAN(x)     - arcus tangent function            arc tan(x)
   ACOT(x)     - arcus cotangent function          arc cot(x)
   ASEC(x)     - arcus secant function             arc sec(x)

   SINH(x)     - hyperbolic sine                   sinh(x)
   COSH(x)     - hyperbolic cosine                 cosh(x)
   TANH(x)     - hyperbolic tangent                tanh(x)

   ASINH(x)    - area hyperbolic sine              ar sinh(x)
   ACOSH(x)    - area hyperbolic cosine            ar cosh(x)
   ATANH(x)    - area hyperbolic tangent           ar tanh(x)

   SQR(x)      - square of x -- faster than x^2    x^2
   SQRT(x)     - square root                       x^(.5)
   ROOT(y,x)   - yth root of x                     x^(1/y)

   LN(x)       - natural logarithm                 ln(x)
   LG(x)       - decadic logarithm                 log10(x)
   LOG(y,x)    - logarithm based y                 log_y(x)
   EXP(x)      - xth power of Euler constant       e^x

   FAC(x)      - faculty of int(x)                 x!

   RAD(x)      - convert degree argument to radians
   DEG(x)      - convert radian argument to degrees

   ABS(x)      - absolute value                    |x|
   SGN(x)      - signum function                   sgn(x)

   MAX(x,y)    - returns the greater of x and y
   MIN(x,y)    - returns the smaller of x and y

 Concluding Notes: Unlike most math programs the operator '/' is interpretd
 as a fraction bar, so everything behind it is interpreted as the denomiator
 unless an addition operator '+' or '-' is encountered. E.g.:

   40/4*5   results in 2
   15/45/9  results in 3
   6/3+2    results in 4 (!)

 There is a conceptional bug in the handling of the '-' operator. Do not
 chain them. Instead write: EXPRESSION - (EXPRESSION + EXPRESSION). Also
 the monadic '-' does not always work the way it should. (-ATOM) usually
 does.

 I won't fix the bugs in the math module, but I am going to completely
 rewrite it, because the concept cannot hold sums, integrals, etc. The
 new module will support integration and symbolic differentiation as
 well.

2 commands
3 general
?command
 The command is the first `token` in a `commandline`. Commands always start
 with a leading sharp sign '#'. Commands aren't case sensitive anymore.There
 must not be any space between the sharp and the command.

 As most commands affect one specific column there is a syntax that is
 common to most commands:

      #command col arg1 ...

 where command is the command name and col is the physical 'column' to affect.
 However, there are some commands that do not affect a specific column or more
 than one, so col represents zero or more `tokens` in the general syntax. The
 same is with the arguments.

 Any `commandfile` must contain the commands `dsize` and `columns`. So read
 the sections concerning those commands carefully.

3 variables
?variables
?global
?enviroment
 There are three types of variables maintained by the program that aren't
 related in any way, may not be mixed up and cannot be processed together
 in any way. Except for the fourth variable that may be build from two
 of the latter.

 The most important are the math processor's variables. Beyond what was said
 in the math processor section, you may use them as data to write to columns
 and `post` them on screen, by means of the equally named command. Specifying
 "$=" in a `data` field makes a maths variable available to the data stream.
 See the `data` section for more information about that. Math variables are
 not cleared by the `reset` command.

 Global variables -- it's just a name, math variables are of course global as
 well -- are used to make the `commandfile` handier and more readable. Global
 variables may replace any number `token` of any command, no care if it is
 a column number, function number, row boundary or whatever. It will be much
 easier to reference columns and functions by name than to keep track of
 hundreds of abstract numbers. The `mark` command even frees you of the task
 to count data rows for your boundaries. See `define`, `queue`, `mark` and
 `undefine` for more information about global variables. Global variables
 are cleared by the `reset` command.

 Terminal enviroment variables are local variables, meaning that their value
 is set per terminal. Two different terminals may carry different values in
 equally name variables. On the other hand they are not cleared by the
 `reset` command. They are used to provide the terminal with additional
 information of how to produce their output. See `set`, `unset` and the
 `terminal` section for more information about enviroment variables.

 Since 1.159 `global strings` are supported. They are useful for two purposes.
 First of all you can use computation results in `terminal enviroment`
 variables, e.g. to add a fit curve to your data plot in the `GNUPLOT`
 terminal. Secondary you can produce a better formatted output on
 runtime.

3 addcol
?addcol
 This command adds a new column to an existing table. Normally you should
 use the `columns` command for that task, but `addcol` might produce more
 readable code or in some cases may cope with memory problems.

 `addcol` must be followed by a number token, which is the physical column
 number to generate. You may pass a name there and it will be assigned the
 highest column number as a global variable. This is basically the way you
 should use this function.

 The second token must be a format identifier as specified in the `columns`
 command.

 Example:

       #addcol LastCol %f
       #row 0
       #using LastCol
       #include "MyData.tst"

 This add a column behind all existing columns, reset the row counter and
 read new data from "MyData.tst" to that column. LastCol would hold the
 number of the new column.

 See also `columns` and `remcol`.

3 applog
?applog
 This command opens a file to store some scratch that you would usually
 write on some sheet for further evaluation during the formulation of
 your data analysis. If the file given with the first token already exists,
 new lines will be appended to it.

 See also `newlog`, `logline` and `logpost`.

3 areg
?areg
 This is the most complex command implemented by now and it will hopefully
 stay forever. `areg` enables you to do a least square approximation to any
 function you like as long as you don't wish to optimize other than two
 parameters. Note that for standard situations e.g. linear or logarithmic
 regressions special commands are implemented so that you won't need this
 command too often. See `lreg` and `logreg` for more information about
 that.

 The syntax is long but logical. The `tokens` are in sequence: x-column,
 y-column, function to match, partial derivation of the function according
 to the first parameter (a), partial derivation of the function according
 to the second parameter (b), estimation for a and estimation for b. While
 all of the latter tokens oblige the following may be omitted: iteration
 depth, start row and finally end row.

 The columns are indicated by their physical column number. The functions
 are indicated by their number as assigned by the `func` command. For both
 groups of numbers as well as for the start and end row, you may of course
 use global variables. The functions are expected to take the parameters:

         y=f(a,b,x)

 in exactly that sequence. Take care that you only pass 3 parametred functions
 matching the above convention.

 The token contents for the estimation values are passed to the maths
 processor, so you may use its math variables or even mathematical expressions
 for the estimation values. Note that `areg` is usually rapidly convergent,
 but false estimations my even crash the iteration.

 If you pass zero to iteration depth or you omitted that token, `areg` will
 work interactively. Thus it will stop after every iteration and show where
 you are. If you're satisfied with the result, you leave the iteration
 pressing the 'Q' key. Any other key will compute the next step.

 Finally you may specify the first and last row of data to include in the
 regression. Row numbers start with 1 and both rows given as boundaries
 will be included. If you specify zero or leave the token empty the maximum
 value will be set, meaning the first row as start or the last row as end.

 The regression results are printed on screen during execution of the
 `commandfile` and are stored to the math variables DPA holding the first
 parameter result, DPB the second and DPSA and DPSB the corresponding
 absolute errors.

 See also `lreg`, `logreg` and `func`.

3 ask
?ask
 Works just like `if` but the condition is given online by the user. The
 rest of the line is promted to the screen and the user is asked to
 enter 'Y' or 'J' for a true condition or 'N' for a false condition. The
 command must further be handeled by `else` and `endif`.

 Note that setting a batch level higher or equal to 10 will always answer
 this command with 'N'.

 Example:

       #ask Shall I do it?
       ! I did it!
       #else
       ! why not?
       #endif

 will print on screen "Shall I do it? (Y/J,N) > " If you type 'N', it will
 post "why not?" or if you type 'Y' or 'J', it will post "I did it!"

 See also `if`, `else`, `endif`, `readvar`, `stop` and `batch`.


3 average
?average
 Followed by the `column` to average, an optional weight column that may be
 left empty and the start and ending row, which as well may be left empty or
 set zero.

 If a weight is specified DataProc computes the average to the weight given
 in the designated column.

 If the startrow `token` is empty or zero the first row is defaulted. The
 analogon is an empty or zero `token` for endrow that is interpreted as the
 last row currently read.

 The results of the average computation is posted on screen during the
 execution of the `commandfile`. The results are also stored to global
 variables that may be accessed by the maths processor. DPMX holds the
 computed mean and DPSMX its absolute error. DPMSX holds the average
 absolute error (standard deviation). Those can be further processed using
 `math`, `post` and `generate`.

 Example:

         #average 3
         ! The average peripherial length is
         #post dpmx*pi

 computes the unweighted average of all entries in column 3. If we assume
 that column 3 holds the results of several attempts to measure the diameter
 of a circle, it finally posts the average length of the peripherial on
 screen.

 See also `mean`, `math`, `generate`, `post` and `lreg`.

3 batch
?batch
 This command sets a so called batch level. The next token must be a
 number token, which is set as batch level. Default is batchlevel 0.
 Any nonzero level will ignore all `stop` commands. A level higher
 or equal to 10 will answer any `ask` command with 'N'.

 See also `ask` and `stop`.

3 build
?build
 This command is used to assemble `global string` variables. It acts much
 like typical print commands in many programming languages. The arguments
 in sequence are: variable name, format string, arguments. Note that the
 argument line is parsed as tokens, thus you will usually have to quote
 the format string or mathematical expressions.

 The format string may contain %d (%ld) what believes the corresponding
 argument to be a `global variable`, %f (%lf) is replaced by the result
 of the given mathematical expression, %s what denotes a pure string or
 a `global string` variable and finally %% producing the percent itself.
 %f (%lf) may be followed by a single digit denoting the number of
 significant cyphers just like the `prec` command. The precision defaults
 to 7.

 Global strings can be evaluated by `build`, `post`, `preamble` and `set`.

 Example:

         #build mystr "The mean pressure is: %f +/- %f hPa" DPMX DPSMX
         #post mystr

 This makes your script's runtime output a little more readable.

3 closelog
?closelog
 Closes a note sheet file that has been opened by `newlog` or `applog`.
 Usually you won't need this command, because DataProc closes all files
 when closing down. Anyhow, if you like to print out your logfile during
 execution, you must close it, before another application may open it
 for further processing.

 Example:

       #applog "mylog.log"
       ; do something
       #closelog
       #dos type mylog.log

 Will print out the log file "mylog.log" that had been generated or modified
 in the same commandfile.

 See also `closetag`, `newlog` and `applog`.

3 closetag
?closetag
 Closes a tag database file that has been opened by `tagfile`. Usually you
 won't need this command as DataProc closes all files when closing down.
 Anyhow, if you want other applications to work with it, before your
 commandfile ends, you may make it accessible by this command. See `closelog`
 for a typical application.

 See also `tagfile` and `closelog`.

3 columns
?columns
 The first `token` is expected to hold the number of columns for the table
 to hold. Since version 1.204 this token is optional and may be omitted. If
 you supply a number of columns it will be used. Note that any format
 identifier beyond that mark will be ignored and if you supply less
 identifiers this will cause an error.

 The next tokens hold format descriptors for the relating column. They are
 just like the printf() C type format descriptors.

      %f    - float number
      %lf   - float number (synonym to %f)
      %d    - integer number
      %s    - character string

 The sequence the columns are declared in makes their physical `column`
 number. E.g.:

      #columns 3 %d %f %s

 would allocate space for three columns and declare the physical column
 number 1 to hold integer numbers, physical column 2 to hold floats and
 physical column 3 to hold strings.

 In addition the logical read and write numbers of the columns are inited
 to the physical column number.

 See also `dsize` and `reset`.

3 contract
?contract
 This command is quite similar to `generate`. But unlike `generate` it
 produces a new column by processing entries of the same column. It could
 be used to compute the deltas between to entries, but that is already done
 by `delta`.

 The first `token` holds the column to generate. Then the start and end row
 follow. Empty tokens there or zeroes are interpreted as boundary values,
 meaning first or last row. The next `token` holds the stepwidth, meaning
 the distance between two rows to process together. So if you have a `data`
 stream holding the occurences of maxima and minima, a stepwidth of 2 would
 serve to compute the e.g. deltas between maxima and minima without mixing
 them up. The next `token` holds the function number to use for processing.
 The following tokens build a column list, so you can use the information of
 more than one column.

 The function arguments are passed in the following manner: Current row of
 the first `column` in the list, current row plus stepwidth of the first
 `column` in the list, current row of the second column in the list and so
 on.

 The destination is written to the destination column at current row plus
 stepwidth. The first stepwidth rows in the destination are thusly marked
 void.

 The row boundaries refer to the current row. Each row between the first and
 last minus stepwidth may become current row, if you select that range.

 Example:

         #func 1 f(u1,u2,i1,i2):=(u2-u1)*(i2-i1)
         #contract 5 0 0 2 1 3 4

 Writes the product of the differences of odd and even row pairs in columns
 3 and 4 to column 5. As the whole range is selected any possible pair in
 forward sequence will be processed. The first two entries in column 5 will
 be marked void.

 See also `func`, `generate` and `delta`.

3 count
?count
 Fills a column with the current row number. An optional function as defined
 by a `func` command may follow. In the latter case the column will hold the
 results of the given function, which is passed the row number as its
 argument.

 Example:

         #func 1 f(x):=10x+3

         #count 1
         #count 2 1

 The first command stores the row number to column 1. The second stores the
 row number multiplied by 10 incremented 3 to column 2.

 See also `func`, `make` and `scale`.

3 define
?define
 Defines a global variable. Global variables may replace `any token` that
 expects numbers. No matter if it be column numbers, function numbers, row
 boundaries, etc. They however aren't related in any way to the math
 processor's variables and thus cannot be used in mathematical contexts.

 The syntax is variable name followed by value. Example:

         #define MyVar 6

 `MyVar` may now replace any number `token` with the value 6. For instance:

         #lreg 1 2 3 myvar

 would do a linear regression of column 2 to 1 weighted by 3, beginning with
 row myvar -- which we defined to be 6. As you probably guess from that,
 global variables -- being `system tokens` -- are not case sensitive.

 See also `undefine`, `queue` and `mark`.

3 delta
?delta
 Works similar to `contract` with the function f(x,y):=y-x as the inbuilt,
 eternal and unchangeable default. Thus the function `token` is missing.
 The first `token` holds the column to generate. The next `token` holds the
 `column` to take the `data` from. If the next `token` holds a '-' sign,
 the function result is negated, resulting in f(x,y):=x-y. If this `token`
 does not hold a '-' it is directly interpreted as the next one, which holds
 the start row and the next the end row. As usual empty or zero tokens mean
 boundary values that is first or last row. The first entry in the
 destination column is marked void.

 See also `contract`.

3 diff
?diff
 Computes the average difference of the entries in a column in a better and
 faster way then a combined `delta` and `average`. It uses the so called
 "far differences method" which produces much better results than the above,
 if you have more than 6 items. The item count however must be even or the
 last item will be ignored.

 The first `token` holds the column to process and the second may optionally
 hold a weight column. By now the latter is simply averaged to compute a mean
 error, which is in turn stored to DPSMX. The mean itself is posted on screen
 and of course stored to the global maths variable DPMX.

 Example:

         #diff 3
         ! Wavelength
         #post 2DPMX

 Suppose column 3 holds the coordinates where zero amplitudes of a standing
 wave have been encountered. `diff` computes the average difference between
 those coordinates that is half the wavelength. The `post` command doubles it
 and puts the result on screen.

 See also `average`, `delta` and `post`.

3 digital
?digital
 Sets the `precision` fixed to the decimal point for the column. The argument
 is the decadic exponent of the last significant cypher.

 Example:
         #digital 2 -2
         #digital 3  1

 will cause the number 123.456 to appear as 123.46 in column 2 and as 120 in
 column 3.

 Beyond that a minimum error is computed for that column which is used in
 averaging, etc. when no error column is specified to avoid meaningless
 results.

 See also `prec` and `average`.

3 dos
?dos
 This command passes whatever it is followed by to the operating system's
 command processor.

 Example:

       #term gnuplot
       #write "MyPlot.dta" "MyPlot.gnu"
       #dos gnuplot MyPlot.gnu

 will imediately plot your results.

3 dsize
?dsize
 Sets the maximum number of rows that all columns of the table may hold. This
 command should preceed the `columns` command.

 If no `dsize` command preceeds the `columns` command, `columns` will default
 the table size to 10 rows. If a table exceeds the given size, DataProc will
 try to enlarge it, by either doubling the current size or adding 100 rows,
 whatever is smaller. Note that this might cause memory fragmentation and
 takes some time, if you have a lot of columns. So if you encouter memory
 problems, try to specify a tight -- but sufficient -- `dsize`.

 `dsize` may now appear anywhere in the commandfile. Note that in previous
 releases this might have caused severe crashes. Anyhow, you cannot shrink
 tables. If you specify a size smaller than the current size, `dsize` will
 be ignored. If you want to save memory during runtime shrinking a table,
 you must `reset`.

 Example:

         #dsize 40

 Sets the maximum number of rows in the data file to 40.

 See also `columns` and `reset`.

3 else
?else
 Will switch to the other condition of an `if` statement. If the last
 `if` already encountered an `else`, it will cancel this `if` and
 `else` refers to the previous nesting level.

 See also `if`, `endif` and `ask`.

3 endif
?endif
 Cancelles an `if` nesting level and returns to non conditional execution.

 See also `if` and `else`.

3 exit
?exit
 Causes DataProc to terminate immediately. Any further data either in
 the current or the parental file or files will be ignored.

 See also `quit`.

3 flushset
?flushset
 Annihilates all assignments to terminal enviroment variables done by
 `storeset` at one time.

 Example:

       ; adapt to german language and EEPIC output
       #include deutEEP
       ; do something
       ; prepare output for a doubly logarithmic regression
       #term latex
       #storeset aID Exponent
       ; produce some output
       ; and return to the "deutEEP" settings
       #flushset

 See also `storeset`, `set` and `reset`.

3 flushvar
?flushvar
 This function clears all global constants. I'm not sure, if you might have
 need for it, because `reset` does this as well, but if you should, here it
 is.

 See also `reset`.

3 func
?func
 The next `token` numbers a function that is explained in the rest of the line
 to the maths processor. By this number `make`, `count` and `scale` reference
 their functions. Note that since 1.231 you may place an undefined global
 variable instead of a defined number token. DataProc will seek for an empty
 slot and define the global with the number it assigned the function.

 The name of the arguments are arbitrary, but the sequence left to right
 corresponds to the left to right sequence how `make` passes the cell values
 of its designated argument columns.

 Example:

         #func 4 c(a,b):=a/b
         #func 5 errc(a,b,sa,sb,c):=c*sqrt((sa/a)^2+(sb/b)^2)

 Computes a value c as the quotient of a and b and its error by the Gauss
 error propagation for potency expressions. The `make` command my look as
 follows:

         #make 5 4 1 3
         #make 6 5 1 2 3 4 5

 This would expect a to be in column 1, its error in 2, b in 3, its error
 in 4 and would generate the quotient c to column 5 and its error to 6.

 See also `make`, `count`, `scale`, `generate` and `math`.

3 generate
?generate
 This command is similar to `make`. The difference is that it does not
 execute during the `data` read, but it does generate a new column immediately
 from the already read table.

 The syntax differs a little from `make`. The first `token` holds the column
 number to generate. The next two tokens hold the start and end row to
 process. Empty tokens or zeroes are interpreted as boundary values, meaning
 first and last row. The next token holds the function number. The value -1
 calls the inbuilt function that results with the row number. The following
 tokens are just as `make` the function arguments.

 Example:

         well we read some data to column 1 and 2 ...

         #lreg 1 2
         #func 1 delta(x,y):=abs(y-DPA*x-DPB)
         #generate 3 0 0 8 1 2

 Column 3 will then hold the absolute difference to the mean straight line.

 See also `make`, `columns` and `func`.

3 globtitle
?globtitle
 globtitle sets a global title to the whole table. The next `token` is
 interpreted as a string which is used to name the table. See the
 `terminal` section for more information about titles.

 See also `title`.

3 group
?group
 The first `token` is expected to hold a group number. If the group number is
 new, a new group is allocated. The following tokens hold a list of columns
 to add to the group. You cannot shrink a group by now.

 Grouping is mainly a formatting advise for the `terminal` driver. Table
 drivers for instance should equip groups with a common title as specified
 by the `name` command.

 Example:
         #group 1 5 6
         #group 1 7
         #group 1 6

 The first command allocates the group 1 and specifies the columns 5 and 6 as
 its members. The second command adds column 7 to the same group that now
 holds the columns 5, 6 and 7. As column 6 is already in the group the third
 command is ignored.

 See also `globtitle`, `title` and `name`.

3 gtitle
?gtitle
 Sets a secondary title to a column. Works just as `title`, but the argument
 string is stored to a different place. This title should be used to e.g.
 name the points in a plot. See the `terminal` section for more information
 about titles.

 Example:

         #gtitle 2 "Counter rates"

 Sets the gtitle to "Counter rates". The gnuplot terminal will thus name the
 data plot "Counter rates".

 See also `title` and `rtitle`.

3 if
?if
 Initiates a conditional block of statements. The rest of the line is passed
 to the maths processor and the result is checked. Any value greater than
 zero denotes TRUE, zero or less FALSE.

 If FALSE, control will not execute any command until `else` or `endif` is
 encountered.

 If TRUE, control will continue execution until `else` is encountered. In
 that case the block will be skipped until `endif` appeares. `else` is of
 course not obligatory.

 See also `else`, `endif` and `ask`.

3 include
?include
 The following `token` is the filename to include. If the file does not open,
 DataProc retries filename.tst and if that fails, too, it issues a "File does
 not exits" error. The include file is expected to be a `commandfile`, just
 like the one calling for `include`. This is especially handy, if you e.g.
 create your data automatically and have to process them just the same
 every time. So you would write a commandfile inputting theese data. Another
 extremely practical usage is to create a datafile by the current script
 and reread the preprocessed data after resetting the table to the same
 script. Well, possibilities seem infinite.

 Whatever is in the file is handeled as if it were part of the original file.
 You may nest `include` as deep you will. As the algorithm is recursive you
 may be well advised to increase stack for extremely deep include nests. On
 other machines like MS-DOS you might have to increase the FILES entry in
 your config.sys, but DataProc's concept has no problem with deep nesting.

 You should however ensure, that you do not create include loops, meaning a
 subfile including one of its parentals. DataProc doesn't check that and will
 crash sooner or later, because of stack leakage or other system odds.
 Depending on the runtime system of the compiler it might even leave files
 open, what may have severe consequences to your filesystem. On some systems
 the medium will become unreadable and will have to be reformatted!!!

3 leave
?leave
 Works similar to `slip` but marks the column to permanently hold misresults
 until a `take` command disables `leave`. This command is used to exclude
 groups of results from a regression, e.g. a non linear area.

 The argument `token` holds the column number to mark as misresults. If the
 column number zero is specified `leave` marks all read columns as misresults.
 Note the the column number refers to the logical read column as assigned by
 `using`.

 Anything that is gained by further processes of a marked entry (e.g. by
 `make` or `generate`) is also marked.

 See also `make`, `using`, `columns`, `slip` and `take`.

3 logline
?logline
 Whatever follows the command will be written to the current logfile. If no
 logfile is opened, the command will be ignored.

 Example:

       #newlog "Important.log"
       #logline Now comes the most important thing ever!

 Will open a new file called "Important.log" overwriting existing an existing
 file and write "Now comes the most important thing ever!" to it.

 See also `applog`, `logpost` and `newlog`.

3 logpost
?logpost
 Works exactly like `post`, but redirects its output to the current logfile.
 If you did not open any logfile, the command is ignored.

 See also `applog`, `logline`, `newlog` and `post`.

3 logreg
?logreg
 Works just like `lreg`, but the first `token` is interpreted as a regression
 mode. Specifying zero makes this command equivalent to `lreg`. One causes
 logarithmic scaling on the x-axis, two on the y-axis and 3 on both.

 This command uses the decadic logarithm. The results that are posted to the
 terminal and set to DPA, DPB, DPSA, DPSB and DPR are the parameters of the
 straight line. It is the user's task and the task of the `terminal` drivers
 to interpret theese results correctly.

 For more information see `lreg`.

3 lreg
?lreg
 Starts a linear regression. The arguments are: x-column, y-column, y-weight,
 startrow and endrow. Only the first two cannot be omitted. An empty or zero
 startrow or endrow sets this value to firstrow or lastrow.

 If a weight column is specified a weighted regression will be done. If not,
 well DataProc does a simple regression.

 The regression results are hooked to the y-column. Output terminals may care
 to output them when outputting the y-column. See the `terminal` section for
 more information.

 Beyond that the variable DPA, DPB, DPSA, DPSB and DPR of the math processor
 are set to the regression results. Where DPA holds ascend of the regression
 straight line, DPSA its error, DPB the y-offset and DPSB the error of the
 latter and DPR the correlation ratio.

 Regression results are monitored to 'stdout' (usually the screen) during
 execution of the `commandfile`.

 Example:

         #lreg 1 3//8 0

 Does an unweighted linear regression of column 3 against column 1 beginning
 with row number 8.

 See also `logreg`, `areg`, `average`, `math` and `post`.

3 make
?make
 This command specifies a function to generate an entry in a column from a
 number of  entries in other columns. All entries belong to the same row.
 Note that `make` won't overwrite valid data. If you want it to, use `void`
 to mark it invalid.

 The first `token` designates the column to generate, the second the function
 number to use (see `func` for more info) and the following tokens are the
 column numbers which the function arguments are to be taken from.

 Example:

         #make 3 1 4 5

 generates the entries of column 3 using the function 1 passing it the
 contents of columns 4 and 5 as arguments each row. The whole thing might
 look as follows:

         #func 1 f(x,y):=x+y
         #using 4 5
         #make 3 1 4 5

 would read the data in the file to columns 4 and 5 and would store their
 sum in column 3.

 The sequence of the `make` commands is irrelevant. Before execution they
 are sorted as necessary that all columns hold valid data before being
 passed as arguments. An error will be issued, if that is not possible.

 See also `func`, `count`, `generate`, `scale`, `undo` and `void`.

3 mark
?mark
 This command is used to `define` a global variable to the current row number.
 It is helpful to make row boundaries easier -- you don't have to count
 anymore -- and more transparent. The only argument is the name of the variable
 to define. Note that you cannot redefine variables. Use `undefine` to free
 them first.

 Example:

         some data here ...

         #mark RegStart

         some data there ...

         #mark RegEnd

         and there may be still some ...

         #lreg 1 2//RegStart RegEnd

 This would cause `lreg` to do an unweighted regression of column 2 to 1
 beginning with the data set following the #mark RegStart and ending with
 the data set following #mark RegEnd -- both inclusive.

 See also `define`, `undefine` and `queue`.

3 math
?math
 Whatever follows this command is directly passed to the maths processor.
 You will usually use it to compute results from regression data and use
 them as constants in new functions or `post` them on screen.

 Example:

         #title 1 diameter
         #unit 1 cm
         #digital 1 -2
         #using 1

         then some data follows

         #average 1

         #math l:=mx*pi
         #math sl:=smx*pi

         ! The average length of the peripherial is
         #post l
         ! with an average error of
         #post sl

 Well, that speaks for itself, doesn't it?

 See also `func` and `post`.

3 mean
?mean
 Works similar to `make`, but you cannot choose an arbitrary function. As
 with `make` the first `token` holds the destination `column`. The next
 tokens hold a list of source columns. The entries of the destination
 columns are filled with the mean of the source columns during the `data`
 reading procedure.

 You can reference the error of the mean using the `smean` command.

 Example:

         #mean 6 1 2 3 4 5
         #smean 7 6

 Stores the mean of columns 1 thru 5 to column 6 and the error of the mean
 to column 7.

 See also `smean`, `make` and `average`.

3 meanof
?meanof
 In many cases you need the average of a well defined set of values for
 further computation. `meanof` frees you of the task of typing it as two
 `math` lines. The syntax is pretty simple. The first token holds the
 name to store the average to, the second the name of its error, then
 a list of maths expressions must follow. Note that you must at least
 supply two expression tokens. Beyond that the total number is arbitrary.

 Example:

       #meanof mr smr r1 "2*r2" 123

 Will compute the average of r1, (2*r2) and 123 and store it to mr. The
 average spot check error.

 See also `mean` and `wmeanof`.

3 merge
?merge
 Adds up to 20 columns to a data column for output purposes. This is used
 to plot more than one column's data in a single diagram by means of the
 "gnuplot" terminal driver. See the `terminal` section for more information
 of how the terminals handle merges.

 Example:

         #merge 2 5 6 7

         well, there might be anything

         #term gnuplot
         #ocol 1 2
         #write "data.dta" "data.gnu"

 This will produce a graph showing the behaviour of the y-data in coloumns
 2 5 6 and 7 to the x-data in column 1. Without the `merge` command only the
 data in column 2 would be taken.

 See also `ocol` and `write`.

3 name
?name
 This command sets a title common to a group. If you are measuring for
 example a set of parameters of a gas at different pressures, you would
 put all measuring series to a common pressure into a group and name it
 according to the pressure used.

 The first `token` holds the group number as specified by a `group` command.
 The second `token` holds the title for the group.

 Example:

         #group 1 1 2 3
         #group 2 4 5 6
         #group 3 7 8 9
         #name 1 "700 mbar"
         #name 2 "800 mbar"
         #name 3 "900 mbar"

 See also `group`, `globtitle` and `title`.

3 newlog
?newlog
 This command opens a file to store some scratch that you would usually
 write on some sheet for further evaluation during the formulation of
 your data analysis. If the file given with the first token already exists,
 it will be overwritten.

 See also `applog`, `logline` and `logpost`.

3 nop
?nop
 No operation. This command does absolutely nothing and is only a kludge
 for debugging purposes.

3 ocol
?ocol
 This command is followed by a list of physical `column` numbers. The
 terminal output is generated from that list in the given sequence. It
 also assigns the logical write numbers of the columns. It is the pendant
 to `using` doing the same for data input.

 Example:

         #term gnuplot
         #ocol 2 9 7
         #write "data.dta" "data.gnu"

 This will output in sequence the columns 2, 9 and 7 to "data.dta". The
 "gnuplot" `terminal` will interpret that as x-data, y-data and y-errors
 and will thus output the corresponding command sequence to "data.gnu".

 See also `term`, `write` and `using`.

3 post
?post
 Whatever stands behind `post` is passed to the maths processor, unless
 the first token is found as a `global string` as defined in `build`.
 The result is then printed on screen during execution of the `commandfile`.

 See `math` for an example. See also `build`, `func`, `lreg` and `average`.

3 preamble
?preamble
 The string following this command is set in front of any terminal output
 via `write`. Internally the string is passed to an fprintf() function,
 so mind the C type character specials. If the first token is found to
 be a `global string` as defined by `build` its contents are used to
 fill the preamble.

 Future versions will support the printout of internal values, but this is
 not yet implemented. Even the C type formatting characters don't seem to
 work properly.

 Example:

         #term gnuplot
         #preamble 'set terminal eepic; set size 1,1.5; set output "data.gph"'

 These commands select the GnuPlot driver as output terminal and the preamble
 in turn inits GnuPlot to output an EEPIC data file to "data.gph" in my
 preferred size.

 See also `term`, `write` and `unamble`.

3 prec
?prec
 Sets the precision for one column or defaults it for all columns if the column
 number 0 is passed. Precision means the number of significant cyphers to
 display. It does not affect the internal computations.

 Example:

         #prec 0 4
         #prec 1 5

 will at first default the precision to 4 significant cyphers for all columns
 and then define the higher precision 5 for column 1. Thus:

   3.0145   will result 3.0145 and in column 1: 3.0145
   3.01458  will result 3.0146 and in column 1: 3.01458  (last digit rounded)
   0.0003   will result 300e-6 and in column 1: 0.0003   (exponents are
                                                          multiples of 3)
   etc.

 See also `digital`.

3 queue
?queue
 Defines a chain of global variables, meaning a number of those with following
 numbers. This is especially useful, if you want to name the coloumns of your
 table. The first entry is expected to hold the value of the first global
 variable in the list. This may of course also be a global variable. The
 following tokens hold an arbitrary list of global variables to `define`. Note
 that you cannot redefine variables, use `undefine` to free them first.

 Example:

         #queue 3 myvar1 myvar2 myvar3

 would set myvar1 to 3, myvar2 to 4 and myvar3 to 5.

 See also `define`, `undefine` and `mark`.

3 quit
?quit
 Causes DataProc to terminate the current file. In the main commandfile
 this will terminate execution -- just like `exit`. In children files
 this will return control to the parent.

 See also `exit` and `include`.

3 read
?read
 Well, for all practical purposes this command became meaningless. You should
 not use it. It redefines the format of logical read columns. The syntax is
 just like the `columns` command, but the identifiers refer to the logical
 read columns and not to physical columns.

 See also `using` and `columns`.

3 readvar
?readvar
 This command enables you to define math variables during runtime. The
 first token holds the variable name, the rest of the line is posted
 to the screen and a prompt is added. Execution waits for a float number
 to be entered and will assign its value to the given variable.

 See also `ask` and `stop`.

3 remcol
?remcol
 Removes a column from the current table. `remcol` must be followed by a
 physical column number. I am not sure, if it is always save to use that
 function, but if you are in trouble with memory space, you might welcome
 this function. Anyhow, if you find crashes that may be caused by `remcol`
 report them and I will see what I can do.

 See also `addcol` and `columns`.

3 reset
?reset
 Starts a new table and thus forgets all columns, regressions, etc. that have
 been done previously. The math processor is however not affected by reset and
 thus all variables, constants and functions are still available.

 reset attempts to free all the memory allocated by table private structures.
 I am not sure if it is always succesful in freeing everything. So if you
 should have memory troubles a reorganization of the `commandfile` may take
 you out of there.

 After processing reset the declarations for `dsize` and `columns` must be
 newly set, because the previous are forgotten and deallocated. Global
 variables are also forgotten. Terminal enviroment variables however are
 not affected by `reset`.

 See also `dsize`, `define`, `set` and `columns`.

3 row
?row
 Sets the row counter to the specified value. This is useful especially
 to read data for different columns from different files as well as to
 partially replace data in an already made table. Be careful using this
 command, you can easily produce "white spots" in your table. The command
 only checks that you do not exceed the maximum size as defined by the
 `dsize` command.

 Example:

       #using 1
       #include "Col1.dta"
       #row 0
       #using 2
       #include "Col2.dta"

 This would read column one from "Col1.dta" and later read the same rows
 but in column two from "Col2.dta".

 See also `using`, `include`, `void`.

3 rtitle
?rtitle
 Sets a secondary title to a column. Works just as `title`, but the argument
 string is stored to a different place. This title should be used to e.g.
 name the regression graph in a plot. See the `terminal` section for more
 information about titles.

 See also `title`, `gtitle` and `runit`.

3 runit
?runit
 Sets a secondary unit to a column. Works just as `unit`, but the argument
 string is stored to a different place. This title should be used for
 hybrid units like in regressions. If no `runit` is specified, the terminals
 should care to create their own hybrids. See the `terminal` section for
 more information about units.

 See also `unit`, `runit`, `gtitle` and `title`.

3 scale
?scale
 Sets a function to scale data before storing it to the table. The function
 must expect one argument of course. `scale` expects the physical column
 number.

 Example:

         #func 1 f(x):=x/milli
         #scale 2 1

 For example the data unit of column 2 may be meters, but you measured
 millimeters and thus copied them to the `data` section. The `scale`
 command will care for that problem and divide all values by milli.

3 set
?set
 Sets a terminal enviroment variable. Please refer to the `terminal` section
 for the meaning of such variables to the terminals. The first `token` is
 expected to be the variable name and whatever follows will be stripped of
 leading and trailing spaces and set to the variable value, unless the first
 token is identified as a `global string` as defined by `build` what sets
 its contents.

 The variable is -- just as the `preamble` -- hooked on the terminal. It will
 not be affected by `reset` and each terminal holds its own set of enviroment
 variables. Thus you must first activate the terminal, you want to change the
 variable in (see `term`). `set` redefines without any warnings, unlike e.g.
 `define`.

 Example:

         #set output myfile.gph
         #set author Lars Hanke

 sets the enviroment variable "output" to "myfile.gph" and "author" to
 "Lars Hanke". Note that the variable name is a system token and thus not
 case sensitive.

 See also `unset` and `term`.

3 slip
?slip
 Marks the following row of the specified column or of all columns, if column
 number 0 is given, a misresult that should not be used for regression or
 averaging. Note that anything computed from that entry e.g. by `make` or
 `generate` will bear the same mark. Also note the the `diff` command will
 ususally produce bad results, if an item is left out.

 The first `token` holds the `column` to affect. If zero is specified it
 affects all column `data` is read to. Other tokens are ignored. `slip` is
 only valid for one single row. It is used to mark one single misresult.
 Note that the column number refers to the logical read column as assigned
 by `using`.

 Example:
         3/3/1
         #slip 2
         4/9/1
         5/5/1

 Obviously the result should have been something like 4/4/1 and 4/9/1 will
 destroy any good regression, so this single line will not be considered by
 a later `lreg` command.

 See also `average`, `lreg`, `diff`, `using`, `columns`, `leave` and `take`.

3 smean
?smean
 This stores the error of the last `mean` command to a new column. The first
 `token` holds the column to store the error to and the second `token` holds
 the column the `mean` was stored to. The whole thing is done during the
 `data` read procedure.

 See also `mean`.

3 stop
?stop
 Promts the rest of the line on screen and waits for any key to be pressed.
 You should use this function to stop a screen scrolling by and watch some
 results.

 Note that a nonzero batch level will cause this command to always continue.

 See also `batch`.

3 storekill
?storekill
 This command is used to locally unset a terminal enviroment variable.
 The syntax and function is equivalent to `unset`, but this command is
 tracked and the variable will be restored with a call to `flushset`.

 See also `flushset`, `storeset` and `unset`.

3 storeset
?storeset
 This command is used to locally set a terminal enviroment variable. The
 syntax is exactly equal to `set`, but this command is tracked and may
 be annihilated by `flushset`. This is useful, if you preset your
 terminals with e.g. a language definition file and want to change some
 settings for a specific output.

3 tagdata
?tagdata
 Produces a clear data entry to a tagfile. Accepts an arbitrary number of
 arguments that may be global strings or math expressions. Note that all
 arguments are tokens. The arguments are evaluated and written to the
 tagfile using the separator given in the first token. If the first token
 is longer than a single character, it will be evaluated as first data entry
 and the separator will default to '/'.

 Example:

       #tagdata & mystr 3*4 2+pi

 will write "(contents of mystr)&12&5.14..." to the tagfile

 See also `tagfile` and `tagpost`.

3 tagfile
?tagfile
 Opens a tagfile or - if you would rather have it - only one tag. A tagfile
 is a textfile containing marks which areas belong to which source files.
 You may post an arbitrary number of lines to each tag and finally read it
 as an input file for DataProc again. As all tags are assured to start with
 a comment, DataProc won't notice them while rereading such a file.

 The syntax is tagfilename in the first and tagname in the second token.
 The tagname may be a global string.

 Calling `tagfile` automatically closes previously opened tagfiles and
 automatically appends new tags, if the given tag is not in the file. It
 also creates new tagfiles, if the given filename is not found.

 Example:

       [something computed the maths: amp amperr DPA DPSA DblDPA DblDPSA]

       #tagfile "MyTags" File1Results
       #build regress1 "%f/%f/%f/%f" amp amperr DPA DPSA
       #build regress2 "%f/%f/%f/%f" 2*amp 2*amperr DblDPA DblDPSA
       #tagpost regress1
       #tagpost regress2

 This would always maintain two data entries in the tagfile "MyTags". You
 may later reread the tagfile as input for another table. This is extremely
 useful, if you have to regress or average the results of several regressions
 or averages of different tables.

 See also `tagpost` and `newlog`.

3 tagpost
?tagpost
 Works exactly like `post`, but the results are written to the current tagfile.
 If no tagfile exists, the command will be ignored.

 See also: `tagfile`, `post` and `logpost`.

3 take
?take
 Removes the misresult tag set by `leave` from a column or from all columns
 if column number zero is specified.

 The argument `token` holds the logical read column to affect.

 See also `leave`.

3 term
?term
 Specifies an output terminal driver. The next `token` is expected to be
 the terminal name. For more information about terminals and output see
 the `terminal` section.

 See also `write`.

3 title
?title
 Sets a title to a specific column. The command has standard syntax. The
 argument `token` holds the string to be used as the column's title. See
 the `terminal` section for more information about titles.

 Example:

         #title 2 Speed

 Sets the title "Speed" to the physical column 2.

 See also `gtitle`, `rtitle`, `unit` and `globtitle`

3 unamble
?unamble
 Deletes a `preamble` from a terminal. Just to get rid of formerly defined
 terminal initialization strings.

 See also `term` and `preamble`.

3 undefine
?undefine
 Cancelles a global variable as defined by `define`, `queue` or `mark`. It
 also frees its memory. You may pass as many variables as you like to free.

 Example:

         #undefine myvar1 myvar2 myvar3

 will free all the 3 variables named. If a variable is not found, `undefine`
 doesn't care.

 See also `define`, `queue` and `mark`.

3 undo
?undo
 Removes a `make` directive from a column and thus stops its automatic
 generation. `undo` removes `mean` and `smean` directives as well.

 Example:

         #make 3 1 4 5

         column 3 is now generated using function 1 with columns 4 and 5
         as function arguments.

         #undo 3

         column 3 won't be generated any longer

 See also `make`, `mean` and `smean`.

3 unit
?unit
 Set the name of the unit in the specific column. In fact this is nothing
 else than another `title`. The terminals care for a handling of the unit
 string as a unit. See the `terminal` section for more information about
 titles and units.

 Example:

         #unit 4 cm
         #unit 2 "\frac{\mbox{m}}{\mbox{s}}"

 Sets the unit of the physical column 4 to cm and of column 2 to a LaTeX
 style command to issue m/s.

 See also `title`.

3 unset
?unset
 Deletes an enviroment variable as defined by `set` from a terminal. See
 `set` and the `terminal` section for more information about enviroment
 variables.

 `unset` may be passed an arbitrary list of enviroment variables. It will
 attempt to free all of them and won't mind, if any of them exist.

 Example:

         #unset author output

 will free both variables "author" and "output".

 See also `set` and `term`.

3 using
?using
 This command assigns the logical read column numbers. using is followed by
 a list of physical column numbers that are in turn assigned logical read
 columns in the given sequence. Columns that do not appear in the list
 will not be read.

 Example:

         #using 3 6 1

 will read the first data `token` into the physical column 3, the second
 `token` to column 6 and the third to column 1. Any further data tokens
 are ignored.

 See the `data` section for  more information about data tokens, etc.

 See also `columns` and `ocol`.

3 void
?void
 This command marks a block of rows in a column as invalid. The syntax is
 in sequence: column to affect, first row to mark void and last row to
 mark void, both inclusively. As the `make` command does not overwrite
 valid columns, `void` can be used to enable remakes in a second read
 pass realized by resetting the row counter using the `row` command.

 See also `row` and `make`.

3 wmeanof
?wmeanof
 This function fulfills the same task as `meanof` but is its weighted
 derivant. The first token holds the variable name to store the average
 to and the second the variable name to store its spot check error. These
 tokens are followed by pairs of maths expressions denoting value and their
 errors each. Note that you must at least supply two pairs -- that is
 six tokens all in all!

 See also `meanof`.

3 write
?write
 Followed by one to four tokens. The first `token` holds the name of the
 output file to write to. The second `token` supplies a secondary filename
 for additional information that is needed by some terminals.

 The last two tokens hold start and endrow, which are both included. Empty
 tokens or zeroes are interpreted as maximum values -- meaning first or
 last row.

 What columns are written depends on their logical write numbers. See `ocol`
 for information about assigning logical write numbers.

 What files are needed, how many columns may be passed and how that is
 interpreted depends on the selected `terminal`. See the `terminal` section
 for more information about that.

 Example:

         #write "data.dta" "data.gnu" 0 8
         #write "dat1.dta" //9

 Writes from the first row to row number 8 to "data.dta" and passes the name
 "data.gnu" for the driver file. Then outputs the rest to "dat1.dta"
 specifying no driver file.

 See also `ocol` and `term`.

2 data
?data
 Any line that cannot be processed as a `command`, `comment` or `post` is
 expected to hold `data`. Data lines are series of tokens. Each `token` is
 read to a logical read column as defined by `using` in the sequence of
 their occurence. An empty or missing `token` causes DataProc to copy the
 last entry again.

 If the first `token` for a column is read empty the default value on startup
 is void.

 Example:

         #using 1 2 3
         #math lv:=2*3.9

         1/2.1
         2/3.5/6
         3/4.7
         4/5.2/3
         5/6.1
         6/6.8
         7/7.2/-
         8/7.5
         9/$=lv

 would result in a table:

         1 2.1 -
         2 3.5 6
         3 4.7 6
         4 5.2 3
         5 6.1 3
         6 6.8 3
         7 7.2 -
         8 7.5 -
         9 7.8 -

 A `token` is marked void, if the last character is a '-' sign. That does
 however not apply to string typed data. It is also possible to read in
 variables from the maths processor as shown in the last line of the above
 example. The magical sequence '$=' causes DataProc to seek for a variable
 that is named as the rest of the `token` and put its value there. This
 feature works with double typed columns as well as with integer typed
 columns.

2 terminals
?terminal
 Together with the regression facilities this is the most useful and complex
 section. DataProc can issue `data` correctly formatted for a number of other
 programs. DataProc generates LaTeX tables (tabular enviroments) using the
 `latex` terminal driver, GNUPLOT data and command files using the `gnuplot`
 terminal and several other terminal e.g. `plain` which produces a printer
 output comes with this version. Read this section carefully, because some
 terminals may handle some situations absolutely different from others.

 All terminals support preambles, which are outputted at the beginnig of the
 secondary outputfile. If there is no secondary file for the specified
 terminal, the preamble goes to the primary. All terminals support backslash
 controls in their preambles.

 See also `commands write` and `commands ocol` for the definition of primary
 and secondary output files, logical write columns, etc. See `commands term`
 to select a terminal for output.

 Terminal names aren't case sensitive anymore, for they are system tokens.

3 backslash controls
?backslash
?character
?control
 Some `terminal enviroment variables` allow non-printable characters, or you
 may wish to set linefeeds etc. into `preambles`. Therefore you may set
 backslash control sequences. They all begin with the `backslash` '\'. The
 definitions of those sequences is similar to the C-language character
 definition. Note that by now you may set '\%' or '%' to result as a percent
 sign. Future versions may however support arguments and then '\%' will
 oblige. The characters following the backslash are not case sensitive. By now
 the following controls are supported.

 `\\` displays the backslash itself

 `\0` is compiled to a null character; ASCII code 0.

 `\b` is compiled to the backspace character.

 `\d` must be followed by exactly three decimal digits. If you've got less,
 place leading zeroes. The sequence will be compiled to the ASCII code
 represented by the decimal number.

 `\f` is compiled to the formfeed character. This will feed the current page
 out of a printer or on most systems clear the screen.

 `\n` is compiled to a linefeed character. On most systems this will not only
 feed, but also cause a carriage return.

 `\r` is compiled to a carriage return character.

 `\t` is compiled to a horizontal tabulator character.

 `\v` is compiled to a vertical tabulator character.

 `\x` must be followed by exactly two hexadecimal digits. If you've got only
 one, place a leading zero. The sequence will be compiled to the ASCII code
 represented by the hexadecimal number.

 `\` must be followed by exactly three octal digits. If you've got less,
 place leading zeroes. The sequence will be compiled to the ASCII code
 represented by the octal number. Note that there is `no` extra identifier
 behind the backslash.

3 dump
?dump
 This terminal is used to auto-generate `include` files for DataProc itself.
 It does not do any data formatting, but the result may be read in right
 away even by the same commandfile e.g. after resetting the current table.
 The `dump` terminal takes care that information about void data or misresults
 won't get lost. It may also preserve titles and enviroment variables including
 preambles.

4 enviroments
?enviroment
?variable
 The very few supported enviroments for `dump` are:

 `Global` will issue a #globtitle command to the destination to preserve the
 current table name.

 `NoSlip` switches off the generation of #slip commands to mark misresults.

 `NoTitle` switches off the generation of titelling commands. Default is to
 issue #title, #gtitle and #unit. Note that you will get misassigned titles,
 if you change the column sequence or start by using when rereading the
 table.

 `Seps` sets the data separator character. Usually a space is used, but you
 may feel like specifying other seperators either to generate output for
 other programs or to get along with string type columns, which are by now
 not quoted. That will change in future. Note that backslash controls are
 supported.

 `TermScan` scans the terminals for their enviroment settings and issues a
 command sequence that sets those variables again when reread. Rereading
 won't however cancel variables that have not been scanned, but set later
 by your commandfile.

3 gnuplot
?gnuplot
 The primary data file holds the data to plot. The secondary file will have
 a sequence of GNUPLOT commands written to. All you will still have to do
 is to start GNUPLOT, type 'load "secondary file"' and you should see your
 graphics on screen.

 The `gnuplot` terminal writes no more than 4 data columns. The first column
 is expected to hold the x-data and the second the y-data. If three columns
 are given the third is interpreted as y-delta error and the errorbars style
 is automatically written to the commandfile.

 If 4 data columns are selected, you will need the inofficial GNUPLOT 3.1 that
 was patched by me. This one supports an errorsquares style meaning x and y
 errorbars. As this is an inofficial feature the terminal output will change
 as soon as an official GNUPLOT supports x errors. Anyhow, the third column
 holds the x errors and the fourth the y errors. The errorsquares style will
 of course be automatically selected.

 Void data is not plotted, but misresults are plotted plain without any
 further marks. Future versions may enable you to switch that off.

 If an `rtitle` is specified for the y-column and a regression had been done
 to it, the regression graph is printed as well. Logarithmic regressions (See
 `commands logreg`) are considered and printed correctly, so that you will
 expect an arced graph in a linear scaled diagram. This graph is titled with
 the `rtitle` string. All regressins are memorized!

 The `globtitle` is set to GNUPLOT's equivalent. The column titles (See
 `commands title`) are used to title the axes and their units are put in
 square brackets. Finally the y-columns `gtitle` is used to label the
 data plot.

 Preambles (see `commands preamble`) are made exactly the first line of the
 GNUPLOT commandfile. You may use them to custom scale the diagram, select
 an output file for GNUPLOT and its terminal for instance. You will also
 find it useful to tell GNUPLOT about scaling the axes e.g. by

         #preamble "set logscale x"

 Grouping is not supported. But you may merge columns (see `commands merge`)
 and get multiple y-data and even regressions printed to the same x-axis.
 Note that the terminal checks, if the x-axes of the various regressions are
 compatible.

4 enviroment
?variable
?enviroment
 Well, there is quite a lot of them for this terminal. You should be able to
 produce any 2-D-plot you like. In the following I will use the Dollar symbol
 '$' to indicate the variable's contents. In any case consult your GNUPLOT
 manual. Here they are in detail:

 `Angles` issues a "set angles $" to the commandfile. This is needed to switch
 between radians and degrees in `polar` plots.

 `AddPlot` contents are added behind a seperating comma behind the `plot`
 command in the command file. This enables you to e.g. plot a theoretical
 curve into your experimental data.

 `Errorbars` forces a 4 column output to be handled with `errorbars` and not
 `errorsquares`. This should always be set by users, who do not work with my
 modified GNUPLOT, unless `errorsquares` becomes GNUPLOT standard in future.

 `FullRange` contents will be verbally inserted at the position of ranges in
 the GNUPlot command file. Note that you have to write the square brackets!

 `Key` issues a "set key $" to the commandfile. This is needed to place the
 graphics key in the plot, thus to get it out of interesting regions.

 `Logscale` issues "set nologscale" if it is left empty. In any other case
 it issues "set logscale $". You may want to use it with `logreg` plots.

 `Output` issues a 'set output "$"' to the commandfile. You should use it as
 soon you are sure the plot fits your needs. It saves a lot of typing.

 `Polar` issues a "set polar" if set. By default "set nopolar" is issued.

 `Range` issues a range after the `plot` command. The brackets are generated,
 so you will only set the pure range stuff. Example:

         #set range 1:8

 will issue:

         plot [1:8] ...

 to the GNUPLOT commandfile. As you see you can only set the x-range by now,
 but if you wish, I may add a more common enviroment variable.

 `Size` issues a "set size $" to the commandfile. You will usually use it
 together with `output` and `term`.

 `Style` adds another data file plot to the `plot` command in the commandfile,
 using the style given in the variable's contents. For Example:

         #set style csplines
         #ocol 1 2
         #write "data.dat" "data.gnu"

 would issue the following `plot` command to "data.gnu":

         plot "data.dat", "data.dat" with csplines

 Note that the `csplines` style -- data interpolation by cubic splines -- is
 also a feature that doesn't come with the standard GNUPLOT, at least not by
 now (GNUPLOT 3.2). But e.g. `lines` would be applicable with standard
 GNUPLOT.

 `term` issues a "set terminal $" to the commandfile. You will usually use that
 together with `output`.

 `tex` adapts the output for TeX style GNUPLOT terminals for a better TeX
 graph formatting. You should always set this, if you intend to use the
 GNUPLOT terminals 'latex', 'pictex', 'epic' or 'eepic'.

 `yrange` works just like `range` but produces a range entry that affects the
 y-range only. Note that `fullrange` overrides `yrange` which in tur overrides
 `range`.

3 latex
?latex
 The traditional terminal and meanwhile the most complex. At first a crude
 overview. The latex terminal produces a LaTeX tabular enviroment for output.
 This is written to the primary output file. The secondary output is ignored.
 The whole table is put into an \fbox{}. The `globtitle` is set as a
 \multicolumn{} stretching across the full table width. If it is missing, the
 field won't be generated. Then group titles are displayed (see `commands
 name`). If two or more columns of the same group are neighboured in their
 logical write numbers, no vertical seperators will be drawn between them and
 the group title will center in that \multicolumn{}.

 The table's head is concluded by the column `title` and `unit`, which are
 printed `title` over `unit` and each column separated by a vertical bar from
 each other. If no units at all are specified the unit line will be
 suppressed. The head is separated by a double horizontal bar from the table
 itself.

 Void entries are printed as '--' medium sized horizontal bars. Misresults
 are printed in italics. The numbers are formatted and rounded as to the
 precision selected in the `commandfile` (see `commands prec` and `commands
 digital`). All entries are centered. Columns and rows are seperated by
 vertical and horizontal bars.

 After the table has concluded, the terminal scans for regressions, averages
 and so on. If an averaging column or a regression's y-column is included in
 the logical write list, all averaging and regression done to each of those
 will be summarized and the results will be posted in new \tabular enviroments
 following the main table. Future Versions may have a feature to influence its
 style or switch it off.

 Regression results are printed as y-title regression type superscript and
 x-title. The superscript by now is German. The release Version will be
 freely configurable. Then the A and B values are printed with errors and
 units. The A unit is built by \frac{y-unit}{x-unit}. If the regressions were
 logarithmic, \log is set before the corresponding unit.

 Avaraging results are printed in a similar manner. Column title, superscript
 (German by now, will be freely configurable) in the first line. The next
 line holds \bar{x} the mean, its error and unit. The last line displys the
 standard deviation with unit after \sigma. The whole thing displays a \pm
 too much (after the standard deviation) by now, but I'll fix that until the
 release version.

 Finally results of 'far differences' made by `commands diff`. Just as
 averaging results, but as no standard deviation is computed there is no
 third line. The mean difference is named \bar{\triangle x}.

4 enviroment
?variable
?enviroment
 The LaTeX terminal supports a whole lot of enviroment variables and is very
 freely configurable. Beyond that I am sure that there will still be need for
 further configuration variables.

 The variable `RegData`, `Mean`, `MeanID`, `Diff` and `DiffID` serve to adapt
 automessages to your native language. Other than in the old versions (1.120
 and below) the default values are now English. I recommend to write an
 `include` file for language adaption.

 Except for the preamble backslash controls seemed not to be necessary and
 thus are not supported by this terminal.

 Here are the enviroment variables in detail. As before I use the Dollar symbol
 '$' to indicate the variable contents.

 `aID` sets the identifier string that preceeds regression slopes. Defaults
 to "a".

 `bID` sets the identifier string that preceeds regression offsets. Defaults
 to "b".

 `Caption` produces a \caption line. You will probably use it together with
 `FloatHead`. The `globtitle` is used as caption and won't be issued anymore
 in the table itself, when `caption` is set. If `caption` holds contents they
 are used as the \label name. Example:

         #globtitle Tabelle
         #set caption mytab

 would produce an output like:

         \caption{\label{mytab} Tabelle}

 `Corr` enables the output of correlation ratios in regression data blocks.

 `CorrID` sets the identifier string that preceeds correlation ratios. Defaults
 to $r$.

 `Diff` sets the default name for "Differencing data", which is posted at the
 beginning of a block containing those data.

 `DiffID` sets a symbol for the differencing data, which is displayed directly
 in front of the numerical value. This string defaults to $\bar{\triangle{x}}$.

 `FloatHead` puts the whole table into a `table` and `center` enviroment as
 most of you will probably do anyway. Well, it saves typing.

 `Mean` sets the default name for "Averaging data", which is posted at the
 beginning of a block containing those data.

 `MeanID` sets a symbol for the averaging data, which is displayed directly
 in front of the numerical value. This string defaults to $\bar{x}$.

 `NoHead` inhibits the standard `tabular` header and titeling. Note that it
 does not influence `caption` or `floathead`.

 `NoReg` inhibits the regression data blocks which are printed following to
 the table itself.

 `RegData` sets the default name for "Regression data", which is posted at the
 beginning of a block containing those data.

 `Size` puts the table -- not the regression data blocks -- into a LaTeX
 enviroment, which receives the name of the variable's contents. You may use
 it for oversized (too wide) tables and select a smaller font for it. For
 Example:

         #set size scriptsize

 would result in the structure:

         optional FloatHead and Caption

         \begin{scriptsize}

         the table itself ...

         \end{scriptsize}

         regression data blocks ...
         optional FloatHead close

 `SuperTab` tells DataProc to use the supertabular enviroment. Note that you
 need the corresponding style file to use this enviroment. As supertabular
 can pagebreak in tables I recommend to use this option for long tables. Note
 also that I cannot place the table into an enclosing \fbox anymore and thus
 it looks a bit different, say more loosely, so for smaller tables you may
 prefer the normal tabular still. `SuperTab` inhibits `NoHead`, but that my
 change, if you request me to do it.

3 plain
?plain
 This terminal was designed for a 7 bit ASCII output e.g. to a printer. You
 may in some respects however select an extended charset and I'm working on
 improving that. Beyond that I plan to write a terminal supporting graphics
 facilities of printers.

 The output format is similar to the `LaTeX` terminal. Please refer to that
 section for detailed information about how the table and regression blocks
 look like. As TeX does, all scaling is done automatically.

 Note that other than the LaTeX terminal, plain supports backslash controls.
 Backslash controls may be specified for any single character definition and
 the `slip` style initer string. Of course like all the other terminals plain
 supports those controls in its preamble.

4 enviroment
?variable
?enviroment
 The enviroment variables differ a little from the `LaTeX` terminal and beyond
 that there are many more of them to customize your output. Indeed I would
 recommend an `include` file for your favorit settings. It may be a good idea
 to select your screen as output file for table previewing and thus have an
 `include` for your terminal as well. Now the enviroments in detail:

 `aID` sets the string preceeding regression slopes in the regression data
 summaries

 `bID` sets the string preceeding regression offsets in the regression data
 summaries

 `Corr` enables output of the correlation ratio in the regression summaries.

 `CorrID` sets the string to preceed correlation ratios. Defaults to r.

 `Cross` the first character of the contents defines the character to use for
 a single horizontal bar crossing a single vertical. Defaults to '+'. On
 MS-DOS systems you may try: #set cross \d197

 `Custom` allows you to custom scale the column widths of the table. The
 variable string is parsed for `tokens`. Each of those tokens is parsed
 for "number$width". This will set the fixed "width" to the table's
 "number" column. Note that `spacing` is added left and right to the
 column width, as well as one character for a vertical bar.

 `Deviation` sets the string to issue in the regression blocks in front of
 the standard deviation result, which is by now only supported by blocks
 generated by `average`. Defaults to "Deviation".

 `Diff` sets the string to issue in the regression blocks for titelling
 data created by `diff`. Defaults to "Difference avaraging data".

 `DiffID` sets the string to issue in front of the `diff` result. Defaults
 to "Mean difference".

 `Feed` sets the number of blank lines in front of the table. Defaults to
 zero.

 `HdCross` sets the character used for a single vertical bar crossing a
 double horizontal. Defaults to '|'. On MS-DOS systems you may try:
 #set hdcross \d216.

 `HdLine` sets the character to use for a double horizontal bar. Defaults
 to '='. On MS-DOS systems you may try: #set hdline \d205

 `HLine` sets the character to use for a single horizontal bar. Defaults
 to '-'. On MS-DOS systems you may try: #set hline \d196

 `InterFeed` issues a formfeed between the table and the regression block
 if set.

 `Margin` sets the number of spaces left of the table and regression blocks.
 Defaults to zero.

 `Mean` sets the string to issue in the regression blocks for titelling
 data created by `average`. Defaults to "Avaraging data".

 `MeanID` sets the string to issue in front of the `average` result. Defaults
 to "Mean".

 `NoReg` suppresses the output of regression blocks if set.

 `NoTable` suppresses the output of the table if set.

 `RegData` sets the string to issue in the regression blocks for titelling
 data created by `lreg` and `logreg`. Defaults to "Regression data".

 `Slip` allows to define a control sequece to mark misresults. The variable
 string is parsed as two `tokens`. The first one switches the slip style on,
 the second off. The string is parsed as a C-type printf() format string.
 Thus you may use e.g. "\x08" to define the ASCII control character number 8.
 As you probably know "\x" introduces a hex number, the backslash alone an
 octal number. You must write "\\" for the backslash itself and "\%" for a
 %-sign. Each of both control strings may not exceed 29 characters. The
 default is nothing.

 `Small` tells the terminal to omit the single horizontal bars. This lets the
 table appear more compact and reduces it to half its length. Try it out.

 `Spacing` sets the number of spaces to place left and right of a table entry.
 Defaults to one.

 `VLine` sets the character to use for a single vertical bar. Defaults to '|'.
 On MS-DOS systems #set vline \d179 is preferable.

2 errors
?error
 Well, I'm not very pleased with the error handling of neither DataProc itself
 nor the math processor. The messages of latter are well defined and need no
 further explanation, but the position marked, where the error should have
 occured, is almost arbitrary. I will think of a better error handler when I
 rewrite the math module.

 The error messages of DataProc itself are typical for compilers. They are not
 too definite and are usually caused by faults lines ahead. I will try to get
 the whole stuff a little more transparent in future versions. At least I will
 cite the errorline and command.

 But here are the massages and what they mean:

   Never mind -- it's a joke: well, ignore it and give me a note, if that
                              occures. The problem is a software bug.

   Invalid column:            the column number given is out of range. In
                              most cases you will have forgotten to define
                              the column (see `commands columns`)

   Unknown command:           tells you that the `command` is unknown. You
                              will supposedly have mispelled it.

   Invalid function:          the function number given was never defined

   $ -> File "$" does not
   open:                      the named file cannot be opened, because of the
                              numbered operating system error.

   Invalid terminal:          the named terminal is not found in the terminal
                              list. You will probably have mispelled it.

   Low memory space:          well, buy some RAM to cope with that.

   Invalid mathematical
   expression:                should be followed by the message produced by
                              the math processor. See my words above of how
                              I think about those.

   Invalid format descriptor: e.g. in `columns` you specified a column format
                              other than %d, %f, %lf, %s

   No data for regression:    In most cases that will mean that the y-column
                              or the wheigt columns are invalid. If they are
                              O.K. check your memory and if that isn't low
                              report it to me, because that may be a bug.
                              This of course applies to `average` and `diff`
                              as well.

   Invalid group:             the group number given was never defined. Define
                              the group or correct the number.

   This should not appear:    well, it's exactly that. There is certainly a
                              software bug.

   Did not expect data here:  should never appear. If so, give me a note.

   Cannot add number tag:     a global constant cannot be set, because either
                              the variable already exists or there is no memory
                              available to store it.

   Unknown number tag:        a number was represented by an unknown global
                              constant

   Global constant expected:  in a #define statement (or similar) an obligatory
                              name of a global variable is missing

   Terminal enviroment
   variable expected:         in a #set or #unset command the variable key is
                              missing

   Terminal enviroment
   variable not parseable:    you will have mispelled backslash controls in an
                              enviroment or preamble. Not that the error is
                              issued by the terminal and thus won't appear
                              when the definition was made.

   Table size exceeded,
   enlarge #dsize:            should no longer appear

   Column already exists:     The column you want to add with #addcol is already
                              part of the table

   Global string $ is not
   parsable:                  The named global string cannot be evaluated. Check
                              your backslash syntax

   The string $ is too long:  A given string is longer than the field it should
                              be written to.

   DOS command "$" failed:    The command you passed to #dos caused an error
                              during execution

   No more #IF opened:        You called #else or #endif, but you did not
                              specify a corresponding #if.

   #IF at line $ is not
   closed:                    You exit a commandfile with an #IF command that
                              was not terminated by #ENDIF

   Identifier expected:       The command assigns a value to some symbolic
                              constant, but its name is missing.

   Obligeous argument
   missing:                   An argument that must be supplied is missing. Only
                              very few functions generate this error. By now it
                              is only #meanof and #wmeanof

   $ writing file "$":        the numbered operation system error occured during
                              output to the named file.

   $ clearing file "$":       the numbered operation system error occured when
                              flushing the buffers of the named file.

   $ closing file "$":        the numbered operation system error occured when
                              closing the named file.

   $ working on file "$":     the numbered operation system error occured during
                              non-I/O operations in the named file.

2 History
? history
 Some words on history. The internals of DataProc have been severly tested
 during my physics practice and should work fine, but without any kind of
 user friendliness. That were the revisions 1.0 to approximately 1.100. The
 rest was spent on more abstract commandfiles and more flexible terminals
 resulting in such as global variables, terminal enviroments and the removal
 of case sensitivity. Performance has increased a little as well, although
 I did not really optimize anything by now. The first version fulfilling
 (hopefully!) this manual is 1.144. As the manual developed from 1.120 there
 might still be some ancient remarks in it. Ignore them, if you find more
 comfortable remarks somewhere else in this manual.

 I added the `areg` command in revision 1.145, but as I already feared, the
 math processor is much too instable and too slow for it. I tried to fit the
 results of a slit experiment to a sinc-function. Using the same data in a
 hardcoded program with the same subroutine worked fine, but the results of
 DataProc are unusable and seem to be slightly divergent. So be very careful
 using the `areg` command. This will -- has to -- improve with the new maths
 processor.

 SAS/C (tm) V6.0 arrived and I recompiled the old stuff. According to new
 compiler warnings I found some errors in the code. Also I removed some bugs
 in the `plain` terminal and made the `GNUPLOT` terminal useable without
 TeX. I fitted the code for compilation on an IBM PC, detecting another set
 of buggy code sequences. Finally I ran the standard Amiga test, meaning
 MungWall and Enforcer and removed DataProc's Enforcer hits. Revision 1.153
 is the product of all those internal bugfixes. DataProc became about 7 kB
 longer but considerably faster.

 Well, the first release should be DataProc 2.0, but I lacked of testers and
 as it all works fine so far, I don't think it to be a big fault to release
 Version 1.153 already.

 Version 1.154 fixed the end vbar of the unit title line of the `plain` terminal.

 Version 1.156 is almost the same code -- apart from a fix of the sgn() function,
 but as I gained confidence in the code I set the optimize flags for the compiler
 and DataProc shrinked about 12 kB and increased a great deal in speed.

 Version 1.157 now supplies the `AddPlot` enviroment variable in the `GNUPLOT`
 terminal to add theoretical functions to experimental data.

 Version 1.162 supplies the `global string` feature and a slightly enhanced
 error report feature. The latter will be one of the major concerns in future.

 Version 1.167 fixes the boundary bug. Previous versions needed the upper
 boundary to be set one higher than told in this manual, namely the first
 excluded row. This bug is now fixed.

 Version 1.169 introduces the enviroment variable `small` in the `plain`
 terminal and fixes a disgusting bug in the `digital` type formatting of
 output. It still does not work how it should, but it at least should not
 produce wrong results anymore. As soon as I'll find some time I'll rewrite
 that routine to ultimately fix that problem.

 Version 1.174 is the promised `digital` fix. The new code might be slightly
 slower, because it now uses floating point maths, but it has a crucial
 difference. It has a chance to work and the first tests seemed very
 positive. Beyond that the new modules have been compiled with SAS/C 6.1
 resulting in even more effective code.

 Version 1.177 is the answer to the reported rounding problems. Why hadn't
 such severe problems been reported before? The trimming routines are now
 completely rewritten and should be stable by now. Anyhow, there might still
 be special cases that I didn't consider, so check the new routines carefully.
 The behaviour has changed a little. Selecting a precision of 4 cyphers will
 always result in a display of 4 significant cyphers, even if the last ones
 are zeroes! Rounding should do fine now.

 Version 1.178 adds the `runit` command and some smaller enhancements in
 configurability.

 Version 1.182 supports the LaTeX `supertabular` enviroment for LaTeX style
 tables.

 Version 1.187 now computes correlation ratios for linear and logarithmic
 regressions, can monitor them to stdout, set them to math variables and
 of course output them to the relevant terminals. Finally I found an Amiga
 specific bug in the stdout management. Here's to you, Tim, now your stdout
 will be sorted, finally!

 Version 1.189 is the answer to a set of guys using this software extensively
 but don't even think of paying for it. Well, this version is equipped with
 a keyfile. It runs perfectly without a keyfile, if you got strong nerves,
 but a valid keyfile will switch off those nasty intermezzos!? Well, see and
 enjoy!

 Version 1.202 is nothing but the (almost) perfection of the shareware
 reminders. You will notice, what it is, while using it. Never mind, pay!

 Just coming back from porting DataProc to MS-DOS. Finally it seems to run
 fine as I thought of carrying my test files with me. I fixed a possible
 bug in the maths processor that the compiler complained about. The handling
 of the Keyfile changed a bit, because fseek() doesn't seem to work how it
 should. In the end the MS-DOS users are the lucky guys, cause the most
 nasty shareware reminder during the execution does not work. Don't be too
 happy, but sure that I'll fix that. Version 1.203 is nothing but the
 recompiled port sources.

 Version 1.205 no longer requires the number token in the `columns` command.
 Beyond that a bug in `columns` had been fixed allowing you to omit the
 `using` command from your scripts. By default `DataProc` will read all
 columns in the given sequence now.

 Version 1.213 now allows to read several columns later. See the command
 `row` that has been added. Beyond that `quit` and `execute` are new and
 prepare an interactive mode. Finally a bug in the plain terminal has been
 fixed. Well, I had some severe crashes, but I couldn't track them down to
 DataProc. Two hours of constant testing with Enforcer and MungWall didn't
 produce any hit ... so I tend to blame it on some other task. Anyhow, if
 you encounter crashes, report them immediately and tell me what other tasks
 are running, libraries installed, etc.

 Version 1.214 is the product of two sleepless nights of debugging and
 the `internal` version now owns a 40kB sized cross debugger. I will now
 go to sleep telling you ... there ain't no more crashes! Ahh, before I
 forget to mention it, there had been a display format bug with void data
 in the plain terminal, this had been fixed during the last two nights as
 well.

 Version 1.221 has the interactive mode added. See documentation for
 special information.

 Version 1.225 is a bugfix compendium. First a bug I most likely have
 implemented myself in 1.213 that prohibited DataProc from making columns,
 which have already been generated by #make, is fixed. Then a typo in the
 GnuPlot terminal was also fixed. And finally an annoying and heavy bug
 in logscaled regression is removed. Note that any results you got from
 logarithmic actions like #logreg used the natural logarithm and NOT the
 decadic as mentioned in the manual. By now this is fixed. Finally weighted
 logarithmic actions had been pure nonsense, this is fixed as well.

 Version 1.227 fixed another typo in the GnuPlot terminal. Logarithmic plots
 should now be perfect.

 Version 1.228 fixed the `polar` enviroment in the gnuplot terminal. It now
 works as it should.

 Version 1.233 is another contribution to user friendliness. First you may
 specify undefined global variables as function numbers in the `func` command
 which will then automatically initialized, second the `dsize` command is
 far less critical than before. See the commands' section for details.

 Version 1.234 added the `addcol` and `remcol` commands. For all practical
 purposes they are obsolete, but you may think them as a contribution to
 the beauty of commandfiles. As I lack of time I had no time to test the
 routines, but for all thats logic `addcol` has to run and `remcol` at least
 should do.

 Version 1.235 is the recompile of the 1.234 MS-DOS port on the Amiga machine.
 The sources are now assured to be identical.

 Version 1.236 is the product of half an hour using the program myself. For
 what do I have testers, if they don't even notice severe uncomfortabilities?
 The GnuPlot terminal now supports a `yrange` enviroment. The LaTeX terminal's
 regression slope and offset identifiers are now configurable. Titles may now
 be fed with global strings. Finally you may now open a logfile to store
 additional information you may need while while writing down your results.

 Version 1.238 fixes a handful of bugs I found in the same half hour. As
 nobody complained, I assume nobody uses the features added since 1.100.

 Version 1.239 adds the precision feature to the `build` command's float
 conversion.

 Version 1.241 fixes a bug that came in with the titles taking global strings
 and an old bug in `mark`. `mark` now really works as described in this manual,
 but it returned a one too small number in older versions.

 Version 1.246 adds the tagfile support and the `dos` command.

 Version 1.247 adds the `if` clause and a lot of interactive stuff. See `ask`,
 `else`, `endif`, `if`, `readvar` and `stop` for more information. Tagfile
 support was enhanced by `tagdata`.

 Version 1.249 adds the enhanced batch support commands `storeset`, `flushset`
 and `flushvar`. The online statisic commands `meanof` and `wmeanof` and
 finally the plain terminal, too, supports the `aID` and `bID` enviroments.

 Version 1.254 has a lot of the keyfile and shareware reminder stuff with
 a completely new concept rewritten. I hope it now works on MS-DOS machines.
 Note that unregistered MS-DOS versions crashed on some systems.

 Version 1.258 adds the `storekill`, `closetag` and `closelog` commands. It
 also fixes some bugs in the terminal enviroment variable storing code. Now
 `input` may be used as a synonym for `include`, because me and some other
 TeX freaks messed them up constantly. Finally the error handlig using
 include files returned the location of the uppermost `include` command. This
 is fixed by now and the correct location will be displayed.

 Version 1.260 now monitors all file I/O errors and traps them. So on Amiga
 systems you should have to press nasty requesters only one time. Originally
 I wanted to find a bug by tracking I/O activity, but the bug is still in
 there. During the installation of the control routines I optimized file I/O
 a little. It seems to me that output got faster even though anything is
 checked by now. Beyond that two new commands are installed. See `batch` and
 `nop` for documentation. See also the chapter `errors`, because a whole lot
 has changed there.

 Version 1.263 is the result of 10 hours of bug hunt and I finally caught it,
 it came in with `flushset` and caused weired error messages.

 Version 1.264 is the recompile of the MS-DOS port. Some smaller bugs in
 the I/O error handling had been resolved. Beyond that the MS-DOS file
 I/O has been completely revised. At least with MS-DOS Version 6 everything
 works all right -- yes the keyfile stuff as well!

 Report errors, suggestions and registrations to:

   MGR, Asgard
   Lars Hanke
   Schlachthofstr. 67
   4690 Herne 2
   Federal republic Germany

   FIDO: Lars Hanke 2:245/5802.22
   UUCP: [e-mail]

 Good luck with this program and happy computing!