regexp-replace and ^

regexp-replace and ^

[Stephane Chazelas]

The way regexp-replace works ^ is going to match in several
places:

$ a=aaa zsh -c 'autoload regexp-replace; regexp-replace a "^a" b; echo $a'
bbb

That should probably be documented. (would also affect the \<, \>, \b, and
various look-ahead operators in PCRE).

Or maybe have a (E) for ERE and (P) for PCRE parameter expansion flags for the
${var//pattern/replacement}.

-- 
Stephane

Re: regexp-replace and ^

[Bart Schaefer]

On Oct 29,  7:08pm, Stephane Chazelas wrote:
}
} Or maybe have a (E) for ERE and (P) for PCRE parameter expansion flags
} for the ${var//pattern/replacement}.

Each of (E) and (P) already means something else.

It ought to be possible to fix regexp-replace.  A leading "^" isn't that
difficult, it just means the loop should only go around once (unless you
expect it to match after embedded newlines).  The tokens that match word
breaks require more thought.  I'm not sure what to make of lookaheads.

Re: regexp-replace and ^

[Stephane Chazelas]

2015-10-29 16:46:35 -0700, Bart Schaefer:
> On Oct 29,  7:08pm, Stephane Chazelas wrote:
> }
> } Or maybe have a (E) for ERE and (P) for PCRE parameter expansion flags
> } for the ${var//pattern/replacement}.
> 
> Each of (E) and (P) already means something else.

Indeed. Bummer.

How about some (#E), (#P)? ksh93 has some ~(E:regexp) ~(P:pcre)
and ~(E)regexp, ~(P)pcre.

Probably a lot more complicated.

A pcre_subst builtin addition to the zsh/pcre module would
probably be easy.

Or a zsubst with several options for pattern/ERE/PCRE, case
insensitive...


> It ought to be possible to fix regexp-replace.  A leading "^" isn't that
> difficult, it just means the loop should only go around once (unless you
> expect it to match after embedded newlines).  The tokens that match word
> breaks require more thought.  I'm not sure what to make of lookaheads.

Well, there's also: (foo|^) and variations where you still want
to loop.

-- 
Stephane
>