Issue 20998: fullmatch isn't matching correctly under re.IGNORECASE

I have the following regular expression:

In [2]: regex = re.compile("ME IS \w+", re.I)

For some reason, when using `fullmatch`, it doesn't match substrings longer than 1 for the '\w+':

In [3]: regex.fullmatch("ME IS L")
Out[3]: <_sre.SRE_Match object; span=(0, 7), match='ME IS L'>

In [4]: regex.fullmatch("me is l")
Out[4]: <_sre.SRE_Match object; span=(0, 7), match='me is l'>

In [5]: regex.fullmatch("ME IS Lucretiel")

In [6]: regex.fullmatch("me is lucretiel")


I have no idea why this is happening. Using `match` works fine:

In [7]: regex.match("ME IS L")
Out[7]: <_sre.SRE_Match object; span=(0, 7), match='ME IS L'>

In [8]: regex.match("ME IS Lucretiel")
Out[8]: <_sre.SRE_Match object; span=(0, 15), match='ME IS Lucretiel'>

In [9]: regex.match("me is lucretiel")
Out[9]: <_sre.SRE_Match object; span=(0, 15), match='me is lucretiel'>

Additionally, using `fullmatch` WITHOUT using the `re.I` flag causes it to work:

In [10]: regex = re.compile("ME IS \w+")

In [11]: regex.fullmatch("ME IS L")
Out[11]: <_sre.SRE_Match object; span=(0, 7), match='ME IS L'>

In [12]: regex.fullmatch("ME IS Lucretiel")
Out[12]: <_sre.SRE_Match object; span=(0, 15), match='ME IS Lucretiel'>

My platform is Ubuntu 12.04, using Python 3.4 installed from Felix Krull's deadsnakes PPA (https://launchpad.net/~fkrull/+archive/deadsnakes).

Here is a patch.

FWIW, here's my own attempt at a patch.

Both patch are almost equivalent (my patch is much simpler but perhaps 
Matthew's approach is more correct in long perspective).

Unfortunately Rietvield doesn't work with Matthew's patch, so I have added my 
comments here.

> -                (!ctx->match_all || ctx->ptr == state->end)) {
> +                ctx->ptr == state->end) {

Why this check is not needed anymore?

> -                    status = SRE(match)(state, pattern + 2*prefix_skip);
> +                    status = SRE(match)(state, pattern + 2*prefix_skip, 
state->match_all);

> -            status = SRE(match)(state, pattern + 2);
> +            status = SRE(match)(state, pattern + 2, state->match_all);

state->match_all is used but it is never initialized.

> > -                (!ctx->match_all || ctx->ptr == state->end)) {
> > +                ctx->ptr == state->end) {
> 
> Why this check is not needed anymore?
> 
After stepping through the code for that regex that fails, I concluded 
that the condition shouldn't depend on ctx->match_all at that point 
after all.

> > -                    status = SRE(match)(state, pattern + 2*prefix_skip);
> > +                    status = SRE(match)(state, pattern + 2*prefix_skip, 
> state->match_all);
> 
> > -            status = SRE(match)(state, pattern + 2);
> > +            status = SRE(match)(state, pattern + 2, state->match_all);
> 
> state->match_all is used but it is never initialized.

I thought I'd initialised it in all the places it's used.

I admit that I find the code a little hard to follow at times... :-(

fullmatch() is not yet implemented on the regex scanner object SRE_Scanner (issue 21002). Is it possible to adapt this patch to fix this omission?

> After stepping through the code for that regex that fails, I concluded
> that the condition shouldn't depend on ctx->match_all at that point
> after all.

Tests are passed without this check. But I'm not sure it is not needed. At 
least without this check the code is not equivalent to the code before adding 
support for fullmatch(). So I prefer to left it as is.

> I thought I'd initialised it in all the places it's used.
> 
> I admit that I find the code a little hard to follow at times... :-(

Indeed, it is initialized in Modules/_sre.c, and it is always 0. Perhaps it 
will be more consistent to get rid of the match_all field in the SRE_STATE 
structure and pass it as argument.

Gareth, this is unrelated issue.

New changeset 6267428afbdb by Serhiy Storchaka in branch '3.4':
Issue #20998: Fixed re.fullmatch() of repeated single character pattern
http://hg.python.org/cpython/rev/6267428afbdb

New changeset bcf64c1c92f6 by Serhiy Storchaka in branch 'default':
Issue #20998: Fixed re.fullmatch() of repeated single character pattern
http://hg.python.org/cpython/rev/bcf64c1c92f6

Thank you Matthew for your contribution.

stage: patch review -> resolved

messages: + msg216019

messages: + msg214287

stage: needs patch -> patch review

versions: + Python 3.5