A Reg Exp Question

Anonymous Monk has asked for the wisdom of the Perl Monks concerning the following question:

Replies are listed 'Best First'.
Re: A Reg Exp Question by swiftone (Curate) on Jun 28, 2000 at 19:08 UTC
This is a job for a full blown parser. You can look into Parse::RecDescent on CPAN. This is because it sounds like you'll want to match even: `print "Bob was $frank \"yes!\" around"."No Way!", $foo, "\n"; #or print "joe" if $bar;` [download] Which a normal regexp will be unable to follow easily. That said, if this is a one-time job with certain rules you can rely on, there are ways to do what you want. For example, if every statement ends with a semi-colon, and the word "print" is never used in strings in the Perl script, you could try matching `/print[^;];/` [download] But be aware that even something as simple as: `if($foo){ print "Bar" } # or $foo= "Ask frank to print the report";` [download] Would mess that up. If you have a situation with many situations, use a real Parser. Note also that you might mess up the logic of a program. See: `print "Foo\n" while &func($bar);` [download] If your program relies on &func($bar) being called several times, you would have eliminated the entire statement if using the simple regexp; Thanks to lhoward for introducing me to Parse::RecDescent*	[reply] [d/l] [select]
Re: A Reg Exp Question by davorg (Chancellor) on Jun 28, 2000 at 19:12 UTC
Presumabley you have an expression something like this: `$text =~ s/print.a_string//g;` [download] The problem is that by default the `.` is greedy. That is to say, it takes up as much of the string as it can. What you need to to change your code to `$text =~ s/print.?a_string//g;` [download] which will make the `.` non-greedy. This is covered in more detail in `perldoc perlre`. -- <http://www.dave.org.uk> European Perl Conference - Sept 22/24 2000 <http://www.yapc.org/Europe/>	[reply] [d/l] [select]
RE: A Reg Exp Question by ZZamboni (Curate) on Jun 28, 2000 at 19:24 UTC
Like swiftone said, for the general case you really want a full-blown parser. But if you can narrow the possibilities, a regex solution may work. My first assumption would be that all the print statements you want to eliminate are contained in a single line. This seems to be the case from your example. In this case, I would suggest processing line-by-line instead of joining the whole thing in a single string and processing that. My second assumption would be that the prints you want to eliminate will have one of the following forms: `print "string" print("string") print HANDLE "string" print(HANDLE "string")` [download] where spacing can be arbitrary, handles are always all-uppercase, and the string does not contain double quotes. Given this, I would do something like the following: `while(<FILE>) { s/print(\s+\|\s$\s)([A-Z]+\s+)?"string"(\s*$)?/1/; print; }` [download] This is untested. Notice that I am substituting the print statement with a 1, so that it does not break things like: `print "string" if (somecond) becomes 1 if (somecond) something if print HANDLE "string" something if 1 etc.` [download] --ZZamboni	[reply] [d/l] [select]
Re: A Reg Exp Question by maverick (Curate) on Jun 28, 2000 at 19:39 UTC
First off, you can pick up the entire file in one bang by setting $/ `open(IN,"somefile.pl"); undef $/; $string = <IN>;` [download] save yourself a while and a join. $string = q{ print "Content-type:text/html\n\n"; print "Some big hunk of code\n"; print("Content-type:text/html","\n","\n"); print ('Content-type: text/html',"\n","\n"); print "Yet another big hunk of code\n"; print "Yet more big hunks of code\n"; print ("Content-type: text/html"); print 'Content-type: text/html',"\n +\n"; print "Some other big hunk of code that does something really cool\n"; print ("Content-type: text/html"); print 'CONTENT-type: text/html' , " +\n", "\n"; print "Some code that ends the program\n"; }; print $string,"\n"; $string =~ s/print[\s'"$]+content-type:\s*text\/html[\\nr\s'"$,]+;// +gi; print "---------------\n"; print $string,"\n"; [download] This covered all the 'normal' ways of printing a line that I could think of. I you run across any more, I'll update the expression. /\/\averick	[reply] [d/l] [select]
RE:(2) A Reg Exp Question by swiftone (Curate) on Jun 28, 2000 at 19:49 UTC
Of course, good programming style dictates that $/ should be made local in a block, so you don't trample any exterior variation: `{ open(IN,"somefile.pl"); local $/; undef $/; $string = <IN>; } #previous version of $/ is now restored` [download]	[reply] [d/l]
RE: RE:(2) A Reg Exp Question by maverick (Curate) on Jun 28, 2000 at 21:14 UTC
oh yeah, sorry. :) I also realized that my expession doesn't handle file handles (props to ZZamboni) I'll fix both here shortly. /\/\averick	[reply]
Re: A Reg Exp Question by chromatic (Archbishop) on Jun 28, 2000 at 22:36 UTC
If you want to dump lines that begin with a specific string, this approach works nicely. Instead of slurping the file (with the $/ technique), apply a quick regex on each line: `# open file, don't do the local $/ trick while (<FILE>) { next if /^\sprint/; next if /^\s#/; push @lines, $_; }` [download] You could, of course, append $_ to a variable called $line if you wished. Feel free to add more small regexes, or modify the existing ones. Sometimes a line-by-line transformation is easier.	[reply] [d/l]
Re: A Reg Exp Question by gaggio (Friar) on Jun 28, 2000 at 20:03 UTC
You could also just replace the actual string by "" using s///. That way, you will print nothing instead of printing something, that is you will DO nothing. Because if I understood it right, the string is always the same and can't be split! Just my 2 cts...	[reply]