Preserve Case in Regexp

[stefan k]

Fellow Monks,

I just ran into the problem of renaming a C++ class throughout many files. The class-name was in camel case (say TheClass) whereas the filenames were lower case (theclass.cc). The bundle came with a Makefile and naturally I wanted to replace the names in there, too.

The renaming of the files was done using mmv.

And for the contents of the files I wanted to use that favorite perl-one-liner:

perl -pi.bak -e 's/theclass/thenewclass/'

And, yes, I could have done this for every possible case, but I wondered whether all that could be done in one line. After reading man perlre and surfing the Monastery I finally found one solution at Re: Regex: Case insensitive search but case sensitive replace.

But, honestly, this can't be it, can it?

I'd appreciate a "I" modifier (or something like that) for regexeps that means "preserve case" very much like the (X)Emacs variable case-replace does it. It would match case insensitive but perform some magic to the replacement in that it would preserve the case of the matched pattern there.

I am well aware that this would be quite hard for replacements that have a different length from the search pattern. Still, even then the cases at some prominent positions (like the first or the last character or characters after a grouping construct, etc) could be figured out.

Is that possible or already existing, or already planned? Am I missing something here?

In expectation of deeper enlightenment and with many thanks...

[Abigail-II]

Is that possible or already existing, or already planned?

Not for perl5, but feel free to submit a patch.

Am I missing something here?

Such a substitution is probably far less common that you might think. Not common enough to use a regex flag and increase the complexity of perl (the interpreter). It would also start endless discussions of what the behaviour would be if the replacement is longer/shorter than the original, and cases where the match is in lower-case, the replacement is in upper-case, and there's no lower-case equivalent for said upper-case. Not to mention the symantics of the /eI combo.

Abigail

[graff]

Given that perl won't do exactly what you want, perhaps a reasonable compromise is to have it provide a fallback that's easy enough.

Suppose you use your favorite text editor just to create a list of the mixed-case name replacements -- e.g. the self-respecting unix guru would do something like this:

$ cat > ClRename.sed
s/OldClass/NewClass/g
s/RedClass/BlueClass/g
s/BadClass/GoodClass/g
...
^D

Now create a version of the list that includes the downcased names along with the original mixed-case ones:

perl -pe 'print lc' ClRename.sed > flAndClRename.sed

The latter output file is now suitable for use as a script (for 'perl -pi.bak' or 'sed -f'). Not the most efficient algorithm, perhaps, but usable.