|
|
 |
 |
 |
 |
Fortran Programming Language
|
 |
 |
 |
 |
|
|
|
|
conversion of floating point constants
I have some f77 codes that I would like to update to f90. This will
include replacing common blocks with modules, putting subroutines
into modules, and so on. One of the tasks that I need to do is to
replace double precision floating point constants with the same
constant specified with an "e" format and appended with a kind
parameter. For example, I would like to change the expression
a = c*1.2345678901234d-9+d
to
a = c*1.2345678901234e-9_wp+d
I need to do this in thousands of lines of code in expressions,
actual arguments to subprograms, parameter statements, and data
statements. This needs to work for all possible floating point
constants (with and without a leading sign, with and without a
decimal point, etc.).
Is there a tool that does this already? If not, does someone have a
perl script that does this? I can get close with a home grown perl
script, but it is tricky to cover all the possible formats while
ignoring all of the things that sort of look like floating point
values but aren't, so I don't want to reinvent a wheel if there is a
well-tested one already out there somewhere.
$.02 -Ron Shepard
Kedit:
arb on $
cha /$d-9+d/$e-9_wp+d/ * *
Skip
On Tue, 22 May 2007 23:58:37 -0500, Ron Shepard <ron-shep @NOSPAM.comcast.net> wrote:
-|include replacing common blocks with modules, putting subroutines
-|into modules, and so on. One of the tasks that I need to do is to
-|replace double precision floating point constants with the same
-|constant specified with an "e" format and appended with a kind
-|parameter. For example, I would like to change the expression
-|
-| a = c*1.2345678901234d-9+d
-|
-|to
-|
-| a = c*1.2345678901234e-9_wp+d
-|
-|I need to do this in thousands of lines of code in expressions,
-|actual arguments to subprograms, parameter statements, and data
-|statements. This needs to work for all possible floating point
-|constants (with and without a leading sign, with and without a
-|decimal point, etc.).
-|
-|Is there a tool that does this already? If not, does someone have a
-|perl script that does this? I can get close with a home grown perl
-|script, but it is tricky to cover all the possible formats while
-|ignoring all of the things that sort of look like floating point
-|values but aren't, so I don't want to reinvent a wheel if there is a
-|well-tested one already out there somewhere.
-|
-|$.02 -Ron Shepard
On May 23, 7:59 am, Herman D. Knoble <SkipKnobleL@SPAMpsu.DOT.edu>
wrote:
> Kedit: > arb on $
> cha /$d-9+d/$e-9_wp+d/ * *
d_new = d_old-9+d
becomes...???
Don't know/have a good tool for the general purpose but brings back
terrible memories of assigning the similar task to a summer intern
during conversion of a huge code from CDC to DEC machine ages ago. A
TECO macro changed all "E" to "D" throughout amongst other things w/o
creating a backup of the file(s) before submitting the job (which took
most of the night shift on the old DEC-10). Needless to say, it
turned out more effective to simply start the project over again
despite two months of effort wasted... :(
In article <oje85358ohf180lpm8rrltehpnn31tp@4ax.com>,
Herman D. Knoble <SkipKnobleL@SPAMpsu.DOT.edu> wrote:
> Kedit: > arb on $
> cha /$d-9+d/$e-9_wp+d/ * *
example. What I want is a tool that appends "_wp" to all floating
point constants, and only to floating point constants, in all the
many ways that floating point constants can be written. I also want
to change the "d" exponents to "e" exponents. Also, if a constant
happens to already end with "_wp" or some other kind parameter
because someone manually edited one already, then I don't want to
append another one. If a variable or expression happens to look
like a floating point constant (e.g. the variable A1234d0 or the
expression A1234d-9), then I don't want to modify those. Since this
is fixed-format f77 code that I'm starting with, there may be
oddball whitespace and continuation line cases that might need to be
handled, although I hope there aren't too many of those, so I could
probably do those by hand if the tool could just point them out to
me.
People have been converting legacy f77 code to f90 for over 15 years
now, so I expect there are well-tested conversion tools, editor
scripts, perl scripts, etc. out there that do this correctly.
> -|I need to do this in thousands of lines of code in expressions,
> -|actual arguments to subprograms, parameter statements, and data
> -|statements. This needs to work for all possible floating point
> -|constants (with and without a leading sign, with and without a
> -|decimal point, etc.).
> -|
> -|Is there a tool that does this already? If not, does someone have a
> -|perl script that does this? I can get close with a home grown perl
> -|script, but it is tricky to cover all the possible formats while
> -|ignoring all of the things that sort of look like floating point
> -|values but aren't, so I don't want to reinvent a wheel if there is a
> -|well-tested one already out there somewhere.
$.02 -Ron Shepard
In case anyone else is interested, here is what I'm using so far.
This is a homegrown perl script with a one-line substitution, and it
doesn't handle all possible cases that might arise in f77, but it
takes care of the common ones.
$.02 -Ron Shepard
=====================
$ cat fp_convert.pl
#!/usr/bin/perl
# usage: fp_convert.pl <infile >outfile
# or fp_convert.pl infile1 [infile2...] >appendfile
# replace all floating point constants written with a d exponent with
# the equivalent e exponent form with trailing _wp.
# matched forms include:
# 123d99
# 123.d99
# 123.456d99
# .123d99
# with optional leading sign and optional exponent sign.
# embedded spaces and continuation line breaks are not allowed.
# 22-May-2007 -Ron Shepard
while (<>) {
s/\b([+-]?(?:\d+(?:\.\d*)?|\.\d+))([dD])([+-]?\d+)/\1e\3_wp/g;
print;
}
|
|
|
|
|
|
Add to del.icio.us |
Digg this |
Stumble it |
Powered by Megasolutions Inc