Re: New escape method proposal (was: Re: Possible solution for special characters in makefile paths)

[Eli Zaretskii]

> From: Paul Smith <address@hidden>
> Cc: address@hidden
> Date: Sun, 13 Apr 2014 15:22:02 -0400
> 
> Someone wants to convert all spaces in the filenames in a variable into
> dashes or periods, but doesn't want to touch spaces that aren't part of
> a filename.  Someone wants to convert the spaces that are NOT part of
> filenames into colons but not touch the spaces that ARE part of
> filenames.  Someone wants to add backslashes before colons that are part
> of a filename in a list of filenames separated by colons, but not touch
> colons that are not part of a filename.
> 
> Some of the above can be handled by using $(foreach ...) to allow
> operation on a single word at a time (the ones dealing with whitespace).
> Of course, that reduces consecutive whitespace into a single space:
> 
>   FILES := $[foo bar]    $[biz baz]<TAB>boz
> 
>   CONVERTED := $(foreach F,$(FILES),$(subst $[ ],--,$F))
> 
> gives "foo--bar biz--baz boz" as the result; multiple spaces and the TAB
> are lost, and we can't treat TABs and SPACEs separately.  Going the
> other way may also lose information:
> 
>   FILES := $[foo bar]    $[biz baz] boz
> 
>   PATH := $(patsubst %:,%,$(foreach F,$(FILES),$F:))
> 
> Here you'd get PATH set to "foo bar:biz baz:boz" but what if you wanted
> this instead: "foo bar::::biz baz:boz" (e.g., convert each individual
> space)?  Can't do it.  And, I don't see any way to handle the last one:
> 
>   PATH = $[foo:bar]:$[biz:baz]:boz
> 
>   # how to make this: foo\:bar:biz\:baz:boz ??
> 
> 
> There is clearly a loss of capability if we don't have a way for make to
> treat the two different "modes" of characters differently.  However, I
> don't know if this is compelling; for example we don't allow you to
> substitute the letter "a" that is part of an escaped filename but not
> letter "a" which is NOT part of an escaped filename: $(subst $[a],-,$X)
> and $(subst a,-,$X) are identical.  Why should "$[:]" be different than
> ":", just because colons are special to make?
> 
> Whitespace IS clearly different, because whitespace is special
> everywhere.  But the $(foreach ...) trick may be sufficient to mitigate
> that.  There are things it can't do, as I show above.  But are these
> important enough?

Thanks for the explanations.

No, I don't think these are important enough to expose internal
representation of embedded spaces to user-land.  Whatever can be done
with $foreach should be enough, and should additional special cases
pop up that we would like to cater to, we could introduce new
functions to allow them.