home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.barnyard.co.uk
/
2015.02.ftp.barnyard.co.uk.tar
/
ftp.barnyard.co.uk
/
cpm
/
walnut-creek-CDROM
/
CPM
/
ZCPR33
/
TCJ
/
TCJ26.MZG
/
TCJ26.MAG
Wrap
Text File
|
2000-06-30
|
25KB
|
467 lines
This is the text of my second article in The Computer Journal. Please
note that there is an error in the discussion of the operation of
flow control commands with colons in the command. The third article,
which has now appeared, discusses the error and sets the record straight.
This file was prepared for the PMATE text editor (hence the odd line
below, which sets the right margin and establishes an automatic
indentation to column 8 for hangling paragraphs). If you want a clean
copy, please subscribe to The Journal by sending $16 for one year or $28
for two years to The Computer Journal, 190 Sullivan Crossroad, Columbia
Falls, MT 59912. Call them at (406) 257-9119. Back issues are
available at $3.50 each ($3.00 for 6 or more). The article below appeared
in issue #26.
R76;I8
The ZSIG Column
Jay Sage
1435 Centre Street
Newton Centre, MA 02159
In my last column I promised that this time I would discuss some ideas
for new Z-System programs and program enhancements. Before turning to that
subject, however, I would like to address another issue.
ZCPR3 on a Floppy-Only System
Many people have said that ZCPR3 is a great operating system but that
one must have a hard disk system to use it. I would like to put that myth
to rest. ZCPR3 CAN BE USED AND USED VERY EFFECTIVELY ON A COMPUTER WITH
FLOPPY DISK DRIVES ONLY. I have Z-System running on six 8-bit computers,
and only one has a hard disk drive (that's the Z-Node machine, which I
almost never get to use myself). Admittedly, ZCPR3 runs much better on a
system with a hard disk drive and/or RAM drive, but everything runs much
better on such a system. Z-System no more requires a hard disk than does
standard CP/M.
Squeezing More Performance Out of a Floppy System
A full-up Z-System with shells, error handlers, a search path, and an
extended command processor (ECP) can make for a lot of disk accesses, and
this can slow operation down. However, there is a simple technique that can
help a great deal and can even benefit a hard disk. To understand the
technique, you need a little background on how the disk operating system
(DOS -- either BDOS, ZRDOS, or P2DOS) works and how data are stored on a
diskette.
Let's take up the second matter first. Data on a diskette are stored
in circular tracks numbered from the outside of the diskette. The first two
tracks (sometimes one, sometimes three) are reserved for the operating
system code and are called "the system tracks." When the computer is first
turned on or when the reset button is pressed, the microprocessor starts
running a program stored in a ROM (read-only memory chip). This program
contains instructions that initiate a process whereby the entire operating
system is read in from these system tracks. Additionally, with each warm
boot, initiated by the user with a control-C command or automatically by
some programs when they are done, the command processor code (and generally
the DOS as well) are reloaded into memory from the system tracks.
With the next track inside the system tracks begins what is called the
directory of the disk. Like the index of a book, the directory has entries
for each file stored on the disk containing such information as the name of
the file and where on the disk the data for that file are stored. The
directory has a fixed size and can accommodate only a fixed number of
entries (that is why if you have many small files, you can sometimes get a
"disk full" message even when there appears to be plenty of room on the
disk). The rest of the disk is devoted to the storage of the data for
files.
Now let's look at what DOS does when it wants to read a file. First it
has to find out where the file is stored, so it moves the drive heads to the
directory. Since the directory, unlike the index of a book, is not sorted,
DOS must start at the beginning and read through the directory until it
finds the entry for the file it wants. Once it knows where to go on the
diskette to find the data, it moves the heads to the proper track and reads
the data. This head movement or seeking, as it is called, can take a long
time, especially with older, slower floppies. Even a fast, modern drive
that steps at a rate of six milliseconds per track will take nearly half a
second to seek from the directory track to the innermost track on an
80-track drive. Clearly files located near the directory track will load
much faster than those located near the inside of the diskette. And files
with directory entries near the beginning of the directory will have a
slight edge over those with their entries at the end.
I think you can now see what we should do to speed things up. But two
further questions face us: 1) which are the files whose loading times are
critical to the performance of a Z-System, and 2) how can one get the
operating system to a) place directory entries for these files at the
beginning of the directory and b) place the data for these files on tracks
near the directory?
The second question is easy to answer. Whenever DOS writes a file out
to disk, it uses the first free entry in the directory and places the data
in the outermost available space on the disk. After a diskette has been
used for a while and files have been written and erased many times,
directory entries and file data can be rather disordered. However, if one
starts with a freshly formatted disk (with only the operating sysgen'ed onto
it), files copied to it will fill the directory and data spaces exactly in
order.
Now comes the question as to which files should be written to the fresh
disk first. The general answer is: the ones that are accessed most often.
In a Z-System one must be careful when trying to answer this question
because there are files that are accessed automatically by the system, and
the user may not even be aware of them. Here are my suggestions for files
to put on the list. The exact order in which they are copied to the disk is
not that critical.
If you are like me and have been known to make typing mistakes when
entering commands, then you will want the ZCPR3 error handler to be called
in so that you can correct the mistake. Put your favorite error handler on
the list (mine is VERROR). If you use an extended command processor (more
on that subject in a future column), you should put it and its associated
files on the list. I use ARUNZ, a program that enables one to store
hundreds of alias scripts in a single, small text file, ALIAS.CMD. Whenever
a command cannot be executed either as a built-in command (in the FCP, CPR,
or RCP) or as a transient (COM file), then ARUNZ is invoked, It tries to
find the given command name in ALIAS.CMD. Since these operations are
performed often, I include ALIAS.CMD and ARUNZ.COM on my list of files.
Another popular extended command processor is LX (or LRUNZ), a program that
extracts programs from a library (COMMAND.LBR) of COM files. If you use LX,
then put LX.COM and COMMAND.LBR on your list.
The next category of programs to consider is shells and their
associated files. I make frequent use of the history shell, HSH, with its
command line history file HSH.VAR. It allows one to perform complete
editing of the command line, just as if one were entering the line with
WordStar. In addition, it lets one recall commands issued previously (even
on a previous day!), edit them as desired, and run them again. What a
pleasure to run a system like this! Another command-line shell, VCED (for
video command line editor) performs similarly, and the choice between them
is a matter of personal preference. Both of them can really slow down the
operation of a floppy-based system if not placed on the diskette as
described above. Other commonly used shells are VFILER (with the VFILER.CMD
macro file) and MENU and VMENU (with their menu files MENU.MNU and
MENU.VMN). If you use any of them, add them to the list.
The last category of files I include on the list is files that I invoke
manually very often and want to see the results of fast. This especially
includes directory programs (SD and XD, for example). I also hate to wait
for my text editor (PMATE), so it is on my list.
Let me finish by mentioning some files that I do not include on the
list. Obviously I don't include files I rarely use, such as the system
segments (SYS.RCP, SYS.ENV, etc.) that are used rarely except at coldboot
time. I also do not include programs that take a fairly long time to run
anyway or which are generally run in batch mode. I put assemblers and
linkers in this category. With them, most of the time is spent actually
computing or reading and writing the files on which they operate. Who cares
if the assembler loads in 2 rather than 5 seconds if the assembly takes 20
seconds or more anyway. I would probably also leave off the list programs
that I tend to spend a lot of time in once they are loaded. Turbo Pascal is
an example. Since it has a built-in editor and keeps the source code and
object code in memory simultaneously, one generally lives inside Turbo while
developing a program. Batch programs, like SUB, SUBMIT, or ZEX, would not
be on my list. Since they initiate a long, time-consuming sequence of
events anyway, why be in such a hurry to get things started.
The mention of ZEX reminds me of one final hint on how to expedite the
use of this speed-up technique. I make a batch file for ZEX (you could use
SUBMIT just as well) with all the copy commands in the order in which I want
the files to be on the new disk. The script for a MAKESYS.ZEX file to be
run from the A drive to make a new system diskette on drive B might look
something like
MCOPY B:=HSH.*
GO B:=ARUNZ.COM,ALIAS.CMD
GO B:=VERROR.COM
GO B:=VFILER.*
GO B:=SD.COM,XD.COM
GO B:=*.* N
GO B1:=1:*.* N
If you leave out a file and have to repeat the process, it is much easier to
edit the batch file than to enter manually all the copy commands again,
especially since you'll probably accidentally skip another file and have to
repeat the procedure still again! The last two lines copy all the rest of
the files in user areas 0 and 1; the 'no replace' option 'N' makes MCOPY
skip over any files already on the destination directory. If you have files
in other user areas, add lines for them, too.
----------
R76;I8
New Program Suggestions
We now turn to the promised subject for this month's column --
suggestions for new programs or program enhancements. I hope some readers
will decide to take a shot at writing these programs and submitting them to
ZSIG. If so, I recommend that you let me know in one of the ways (including
writing to me) described in my column in the last issue of The Computer
Journal. That way we won't have two versions of a program that have to be
integrated. Those of you who perhaps do not have the assembly language
programming knowledge can still contribute by sending in further suggestions
and ideas. As I said in the last column, the real ingenuity in software is
not so much in the actual coding as in the conceiving of new program ideas.
SETPATH
The path feature of ZCPR3, which allows a COM file to be loaded
automatically from a drive and user area other than the one from which the
command was issued, is probably the most used (and least noticed) feature.
The utility used to support this capability, PATH.COM, operates in only two
ways. When invoked with no command tail, it displays the search path
currently in effect, listing each path element symbolically (possibly with
'$' characters for current drive or user), in DU format, and in named
directory format. When invoked with a command tail, the tail is interpreted
as a new search path, replacing the old one.
I have often wanted to make simple changes to the path, such as adding
a new element at the beginning or end of the path. Later I want to remove
it. With the present PATH command one has to enter a complete new path
specification each time. I would propose a new program, which might be
called SETPATH, that would offer more flexible options. The syntax for
SETPATH would be as follows:
SETPATH [/option element-list] [/option element-list] ...
The following option specifiers would be recognized:
C clear existing path and append elements listed
A append path element list to existing path
P prefix path element list to beginning of existing path
D or R delete (remove) listed elements from existing path
After the new path is built and loaded, the final path would be displayed.
Option 'C' would be the default option if no option is given explicitly, and
thus SETPATH would work the same as PATH when used the same way. Just to
make sure that the syntax is clear, here is an example SETPATH command:
SETPATH /A NEW1 NEW2 /D OLD2 /P NEW3
If the old path had been "OLD1 OLD2", then the new path would be "NEW3 OLD1
NEW1 NEW2".
There are some technical details to consider. The new path should be
built up in a temporary buffer and its length checked only at the end, since
additions made before deletions might make the path temporarily longer than
allowed. If the final path is too long, one should probably leave the path
as it was, display an error message, and ring the console bell. One might
also want to set the ZCPR3 program error flag (and clear it when the command
succeeds). For security reasons, named path elements must be checked for
passwords, and drive/user path elements must be checked against the maximum
drive and user values specified in the environment. Naturally, like PATH,
the entire utility should work only if the wheel byte is on.
To be complete and thorough, one might want to recognize a leading pair
of slash characters not as an option but as a single slash in a path element
name. Otherwise one would have no way to use a named directory whose name
begins with a slash. The first version of SETPATH could omit this feature,
since it is not critical. A command tail with '//' or '/' only should
display a built-in help message.
There is an issue with the 'D' and 'R' options. In some cases an
element may appear more than once in a path. My default path, for example,
is "SYS ASM MODEM SYS". SYS appears both at the beginning, so that it is
searched first, and at the end, so that programs that access the root
element will also use SYS. My choice would be to make the 'D' and 'R'
options delete only one occurrence of an element. Perhaps 'D' could delete
starting at the beginning of the path and 'R' could remove starting from the
end of the path. Usually, of course, the path element would appear only
once, and both options would give the same result. I'm not completely sure
what should be done when the element to be deleted is not found. The path
should still be built, and an error message should probably be displayed,
but the ZCPR3 program error flag should probably not be set.
SETPATH, as described here, would be fairly easy to write, since it
could borrow most of its code from the present PATH command. An alternative
enhanced PATH command, a video program which might be called VPATH, would
display the path in a full-screen format and allow the user to edit the
path, inserting and deleting elements using cursor controls and editing
commands. My present feeling is that such a command would be overkill.
SETPATH would be adequate and would have the advantage that, not requiring
interactive input, it could be run from batch files and aliases.
R76;I8
SETNDR
The named directory facility in ZCPR3 can be very convenient. I find
it handy to put all my assembly language tools in a directory called ASM and
my modem related programs in a directory called MODEM. It is much easier to
remember the names than the particular user numbers chosen. Some programs
make automatic use of named directories. The HELP program, for example,
looks in a directory named HELP for the file containing the help
information. Echelon's disk cataloguing program DISCAT uses a directory
named CAT to keep the catalogues and a directory BACKUP to determine the
drive holding the diskette to be catalogued.
A program similar in concept to SETPATH, which might be called SETNDR,
would be useful for making simple manual additions, deletions, or changes to
the named directory register (NDR). For example, as mentioned above, DISCAT
catalogues diskettes in the drive containing a directory named BACKUP. In
order to make DISCAT work with a different drive, one would normally have to
have or make another NDR file and load it with LDR. SETNDR would be much
easier to use.
SETNDR could be a very simple, short program with the following syntax:
SETNDR DU: remove name associated with area DU:
SETNDR DIR: remove directory with name DIR
SETNDR DU:NAME assign NAME to area DU:
SETNDR DIR:NAME change directory DIR to NAME
A second name on the command line could be used to assign a password to the
directory.
SETNDR DU:NAME PASS define new name and password
SETNDR DIR:NAME PASS assign new name and password
Before assigning a name to a drive/user area, the program should check to
make sure that the name is not already assigned to a different DU. If it
is, there are three reasonable possibilities.
1. the program could display an error message and refuse to make the
change;
2. the program could delete the existing assignment, provided it is to
a user-accessible area (i.e., not password protected or beyond the
maximum drive and user values specified by the environment); or
3. the program could report the situation to the user and prompt as to
whether the existing assignment should be deleted (again provided
the assignment to be deleted is to a user-accessible area).
My preference is the second choice, but a message should be displayed
reporting the assignment that has been deleted in case it was unintentional.
Like SETPATH, this program should be fairly easy to code. The parser
built into ZCPR3 would do almost all of the work required to interpret the
command line. The programmer would only have to detect the absence of any
command tail or one with only '//'. In these cases a built-in help screen
should be displayed. Most security issues would also be handled by the
ZCPR3 command-line parser. However, the program code should check the wheel
byte and prohibit changes if it is not set.
If the ZCPR3 parser is used, some possibly erroneous input would pass
undetected. For example, the command lines
SETNDR DU:NEWNAME DIR:
SETNDR DU:NEWNAME U:
would ignore the password field (since the parser would see no file name).
Also, a command like
SETNDR NEWNAME
will assign the new name to the currently logged-in area. The latter is
probably acceptable (perhaps desirable). In any case, if one wants to
control these cases, the code would have to double check the actual command-
line tail and not rely only on the default file control blocks.
There is already a utility called LDSK (load disk) for automatically
setting up named directories after one switches diskettes in a floppy drive.
If a user area on the diskette contains a file whose name begins with a
hyphen, then LDSK assigns that name (not including the hyphen) to that area.
LDSK might have some code one could borrow for SETNDR.
R76;I8
Enhanced IF.COM
Conditional command processing, though rarely invoked directly in user-
entered commands, gives ZCPR3 alias and batch scripts tremendous power and
flexibility. Scripts can test such conditions as:
whether a previously run program has set or cleared the program
error flag;
whether a parameter value has been specified or omitted in an alias
invocation;
whether a specified file exists;
whether the file not only exists but has non-zero length;
whether one of the ten user registers contains a specified value;
whether the wheel byte is set or clear; or
whether or not a terminal definition (TCAP) is loaded.
Some of the conditional testing is performed with the ZCPR3 resident
FCP (flow command package). Optionally, the FCP can pass on conditional
testing to the transient program IF.COM. The user can also force invocation
of the transient IF processor by including a directory specification with
the command, such as "A0:IF..." or even just ":IF...". Since IF.COM does
not have to be resident and permanently reduce the memory available for
program operation, it can be a bigger and more capable program. I will
mention here some enhancements that would be useful in IF.COM.
Current IF processing allows one to determine the presence or absence
of a parameter with the NULL option. Howard Goldstein has requested an
ambiguity option:
IF AMBIG AFN
This would allow an alias script to determine if a parameter passed to it
was ambiguous, so that an error message could be echoed instead of passing
an ambiguous file specification to a program that requires an unambiguous
file name.
The current IF processing allows one to determine the existence and
non-zero size of files with the EXIST and EMPTY options. It would be useful
to be able to test the attributes of files, as in
IF ARCHIVE AFN do files have archive bit set
IF RO AFN are files read-only
IF RW AFN are files read-write
IF WP AFN are file wheel protected
IF SYS AFN are files of system (SYS) type
IF DIR AFN are files of directory (DIR) type
The current version of IF.COM allows a test of the form
IF AFN1=AFN2
which compares two possibly ambiguous file specifications. It can be used
fairly generally to compare command line tokens, as in the following
VFILER.CMD macro script:
IF %PT=LBR;LDIR %$;ELSE;ECHO NOT LBR FILE;FI
VFILER replaces the parameter %PT with the type of the file currently
pointed to. If the pointed-to file is a library, its directory will be
displayed; otherwise an error message will be echoed.
There have been a number of times when I have wanted to test
inequalities. ARUNZ alias scripts can read values from the user registers
and from memory locations. Sometimes one would like to test for values that
are less than or greater than some other value. Thus it would be handy to
have tests other than just equality. Equality is the easiest to code
because the ZCPR3 command line parser already handles the equal sign (since
it is used in commands such as REN and CP). For the extended comparisons, I
would propose a syntax of the form
IF TOKEN1 TOKEN2 RELATION
The first two tokens (words) on the command line will be parsed by the ZCPR3
command processor and placed into the two default file control blocks at 5Ch
and 6Ch. The IF.COM code would have to scan the command tail saved starting
at memory location 80h to see if there is a third token. The following
tokens could be recognized:
EQ or = token1 same as token2
NE or <> token1 not same as token2
LT or < token1 less than token2
LE or <= token1 less than or same as token2
GT or > token1 greater than token2
GE or >= token1 greater than or same as token2
Any wildcard characters in either token would be taken as equal to any
corresponding character in the other token. Thus "ABC*" would be equal to
"?BC", since the "A" in the first position matches the "?" in the second
token and the "*" in the first token matches all the blank spaces in the
last five character positions in the second token.
I think these ideas should be enough to keep you busy for a while! I
have several possible subjects for my next column but will await your
response before deciding what to cover. Please send in your suggestions.