NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/unix/question/9634

Text File | 1992-07-31 | 3.4 KB | 108 lines

Newsgroups: comp.unix.questions
Path: sparky!uunet!cs.utexas.edu!zaphod.mps.ohio-state.edu!news.acns.nwu.edu!casbah.acns.nwu.edu!navarra
From: [e-mail] (John Navarra)
Subject: Using sed to convert numbers (interesting)
Message-ID: <[e-mail]>
Sender: [e-mail] (Usenet on news.acns)
Organization: Northwestern University, Evanston Illinois.
Date: Fri, 31 Jul 1992 12:11:18 GMT
Lines: 97

This is an interesting little problem I am working on. I am trying to
convert numbers from scientific notation to regular (decimal) notation.
I am using sed to do this. So far, this is what I have:

[casbah:411] ~/bin -> cat Etodec
sed -n '
s/\([0-9]\)\.\([0-9]\)\([0-9]*\)E+01/\1\2.\3/p
s/\([0-9]\)\.\([0-9][0-9]\)\([0-9]*\)E+02/\1\2.\3/p
s/\([0-9]\)\.\([0-9]*\)E-01/.\1\2/p
s/\([0-9]\)\.\([0-9]*\)E-02/.0\1\2/p' $*
[casbah:411] ~/bin ->

and I have the file with scientific numbers:
[casbah:414] /tmp/navarra -> cat sci.numbers
1.2345E-01
-1.2345E-01
1.2345E+01
-1.2345E+01
1.2345E-02
-1.2345E-02
1.2345E+02
-1.2345E+02

Notice first of all that all the numbers are of the form:
[-]X.XXXXXE[+-]XX

In particular, the fact that there is only one number before the decimal
is crucial in the substitutions.

So far, Etodec (scientific 'E' to 'decimal') is working fine:
[casbah:415] /tmp/navarra -> Etodec sci.numbers
.12345
-.12345
12.345
-12.345
.012345
-.012345
123.45
-123.45

However, just to spark some discussion on this, I have two questions.

1) for you sed gurus, I don't understand WHY my substitutions are working
   for negative numbers. For instance, the line

s/\([0-9]\)\.\([0-9]*\)E-01/.\1\2/p

   looks for numbers of the form   '1.2345E-01' or '-1.2345E-01'. However,
   I am confused as to how the hold buffer is working. The first thing
   I look for is the first number before the decimal (placing it in the
   first hold buffer), followed by a '.' Then the rest of the numbers
   are placed in the second hold buffer. What I don't understand is
   why the replacement pattern '.\1\2' doesn't produce something like:

   .-12345

   And similarly, for the case of E+01, why I don't get something like:

   1-2.345

   If I try the line:
s/\([0-9]\)\.\([0-9]*\)E-01//p

   I get -'s on the matched lines.
   How does sed know to put the '-' back at the beginning of the line?


2) Secondly, as you can probably tell, as the numbers get bigger or
   smaller, the substitution lines will as well. I wrote the line
   to limit the number of hold buffers to three for positive numbers
   and two for negative numbers. However as the numbers get bigger
   or smaller, here is what those lines will look like:

positive
s/\([0-9]\)\.\([0-9]\)\([0-9]*\)E+01/\1\2.\3/p
s/\([0-9]\)\.\([0-9][0-9]\)\([0-9]*\)E+02/\1\2.\3/p
s/\([0-9]\)\.\([0-9][0-9][0-9]\)\([0-9]*\)E+03/\1\2.\3/p
...
s/\([0-9]\)\.\([0-9]...[0-9]...[0-9][0-9]\)\([0-9]*\)E+XX/\1\2.\3/p

negative
s/\([0-9]\)\.\([0-9]*\)E-01/.\1\2/p
s/\([0-9]\)\.\([0-9]*\)E-02/.0\1\2/p'
s/\([0-9]\)\.\([0-9]*\)E-03/.00\1\2/p'
...
s/\([0-9]\)\.\([0-9]*\)E-XX/.0000.....0....\1\2/p'

   What I need is a variable which will scan what the value of 'XX' is
in E[+-]XX and for numbers > 1 (E+XX) , write out 'XX' number of '[0-9]'
strings in the pattern, and for numbers < 1 (E-XX), write out XX-1 zeroes
in the replacement string. Any suggestions on how to do this?

    There is also one more problem I noticed for postive numbers
which I will bring up if a discussion insues.

Any suggestions are appreciated,
-tms