'/e' regexp modifier is not recognized by re pragma #11935
Comments
From glitchmr@myopera.comCreated by glitchmr@myopera.comWhen '/e' modifier is used with 're' module, Perl complains about unknown C:\Users\Konrad>perl -e "use re '/e'" Perl Info |
From @jkeenanOn Sat Feb 04 04:05:02 2012, glitchmr@myopera.com wrote:
Agreed. I couldn't find any discussion of '/e' in 'perldoc re'
|
|
The RT System itself - Status changed from 'new' to 'open' |
From @ikegamiOn Sat Feb 04 04:05:02 2012, glitchmr@myopera.com wrote:
That message is accurate.
"e" is not a regular expression flag. It is a substitution operator flag. |
From @ikegamiOn Sat Feb 04 16:15:49 2012, ikegami@adaelis.com wrote:
Also note that "e" is not mentioned in perlre. The regex flags are (as (I find this very odd that p, g and c are regex flags instead of match |
From tchrist@perl.com"Eric Brine via RT" <perlbug-followup@perl.org> wrote
That's incorrect, and potentially misleading. With blank lines removed for brevity, these are =head2 Pattern Modifiers That's all. Those apply to patterns. These flags, however, apply to the match operator, which is different: =head2 The m// Operator (Matching) Here are the s/// flags: =head2 The s/// Operator (Substitution) And tr/// has its own as well: =begin table picture tr/// Modifiers See? Four distinct sets of flags.
You should find it odd, because they aren't. $ perl -E 'use re "/e"; say "ok"' $ perl -E 'use re "/g"; say "ok"' $ perl -E 'use re "/c"; say "ok"' $ perl -E 'use re "/p"; say "ok"' What's going on with the first three is that it isn't properly dying. $ perl -Mwarnings=FATAL,all -E 'use re "/g"; say "ok"' That's a bug. What's going on with the last one is more subtle. /p can be Anyway, probably perlre should be clearer (read: more correct) --tom |
From @ikegamiOn Sat, Feb 4, 2012 at 8:24 PM, Tom Christiansen <tchrist@perl.com> wrote:
You removed "g" and "c", but they are accepted in patterns.
|
From @ikegamiOn Sat, Feb 4, 2012 at 8:24 PM, Tom Christiansen <tchrist@perl.com> wrote:
With blank lines removed for brevity, these are the only pattern modifiers: You removed "g" and "c", but they are accepted in patterns:
I don't know why that is. If anything is misleading is that they are listed as pattern modifiers - Eric |
From tchrist@perl.comEric Brine <ikegami@adaelis.com> wrote
I removed them because they are non-sensical, and I consider it a bug
Something else that's misleading is having to look in two different --tom |
From @demerphqOn 5 February 2012 01:19, Eric Brine via RT <perlbug-followup@perl.org> wrote:
Just to note, perlre does not decide this. regexp.h does.
Huh?! /p relates to how we manage /g relates to how match /c relates to how we match. None of them have anything to do with substitution. On the other hand /e does, as does /ee Yves -- |
From @demerphqOn 5 February 2012 05:48, Eric Brine <ikegami@adaelis.com> wrote:
Personally I consider "ignored and produce warnings" to be quite Yves -- |
From @demerphqOn 5 February 2012 11:56, Tom Christiansen <tchrist@perl.com> wrote:
Why? You think throwing a fatal is preferred? They do nothing, they
Yes agreed. Yves -- |
From tchrist@perl.com
How can this not be a bug? $ perl -E 'no warnings; use re "/g"; say "gosh that was ok, then, eh?"' $ perl -E 'use warnings FATAL => "all"; use re "/g"; say "gosh that was ok, then, eh?"' $ perl -E 'use warnings FATAL => "all"; use re "/WTF"; say "gosh that was ok, then, eh?"' I can't disable them. Yes, I want a fatal. I certainly want to be *able* to have a fatal. I think they should be a fatal. They most definitely shouldn't be noise. --tom |
From @demerphqOn 5 February 2012 02:24, Tom Christiansen <tchrist@perl.com> wrote:
Please Tom, it is very difficult to discuss things that are a mass of
....
....
tr/// has nothing to do with the regex engine, IMO it has no place in
I personally would analyse modifiers in a very different way. As far as I am concerned there are only three types of regex modifier. 1. Modifiers which change the semantics of how a pattern is interpreted /m /s /i /u /d /l /a 2. Modifiers which change the behavior of the matching process or how /x /o /g /c /p 3. Modifiers which change the behavior of the substitution process. /e /ee /r
I don't really understand this analysis. This is a new feature, what
This one is pretty weird. It makes some sense, but I cant help but
Is it, it seems like it warns, which would be a deliberate action, so
Correct.
Erm, maybe, but personally I would not document it the way you did. It cheers -- |
From @demerphqOn 5 February 2012 18:44, Tom Christiansen <tchrist@perl.com> wrote:
My personal feeling on this is that it matches the general tradition I did not implement the feature, but considering the warning was
So then fix your code.
Well, I personally could see it either way, and am inclined slightly
Warnings are noise? Yves -- |
From tchrist@perl.com
Warnings that are not subject to the warnings pragma are not I cannot enable them. Yes, they're noise. --tom |
From @demerphqOn 5 February 2012 19:01, Tom Christiansen <tchrist@perl.com> wrote:
Oh, I see.
Well the fact that you cant do this ddes seem like a bug, I Yves -- |
From tchrist@perl.comdemerphq <demerphq@gmail.com> wrote
Guilty. Sorry. Yes, you're right. Am rushed.
I don't want people slingshotting around looking for where the devil the The four (or five) pick-your-own-quotes pseudofunctions qr//, m//, s///, Also, all four (or five) of the flaggy PYOQ pseudofuncs can be found $ perl -E 'say "frobnitz" =~ qr/([AEIOUWY])/i' And s///'s /r modifier does exactly the same thing as tr///'s /r % perl -E 'say "frobnitz" =~ y/aeiouy/yuoiea/r' But I do understand that tr/// doesn't belong in a section on pattern The tr/// transliteration operator does not interpolate variables; it
Wait, you mean the RHS not the LHS, right? That's why you But why mention /ee distinct from /e? It's not special.
I figure people looking for what modifiers they can apply to The /dual things are so strange. They're a mutually-dependent We could use a better way to do things. I'd like to figure out a way to make I find qr/PATTERN/msixpodual to be way out there, and adding in Plus now those aren't even listed in the same order as the m//flags are, Might as well use /mxyzptlk after all, eh? --tom |
From @khwilliamsonOn 02/05/2012 11:27 AM, Tom Christiansen wrote:
Patches welcome! |
From @demerphqOn 5 February 2012 19:27, Tom Christiansen <tchrist@perl.com> wrote:
These things happen. But please *do* try.
Well fair enough. You are talking about a book tho, and I was more
They dont change how _substitution_ is performed. They change how the You see, the way you are analysing overlooks the following (to me) Any modifier that affects a qr// is legal in an m// Or in other words: qr// compiles a pattern, m// compiles a pattern and Your analysis seesm to ignore this heirarchy in a way that I do not
Yes I know, but /ee and /eeee are synonyms. And /ee *is* distinct from
But they are collected in a way that is not useful.
Not much choixe there.
I dont understand your point here really. We have certain modifiers cheers, -- |
From @tamiasOn Sun, Feb 05, 2012 at 07:46:55PM +0100, demerphq wrote:
/ee and /eeee are not synonyms. /ee performs two evals; /eeee performs Ronald |
From @ikegamiOn Sun, Feb 5, 2012 at 12:22 PM, demerphq <demerphq@gmail.com> wrote:
Perl disagrees.
|
From @demerphqOn 5 February 2012 23:12, Eric Brine <ikegami@adaelis.com> wrote:
No it does not. I said "relates to how we match", which means "the m// modifier". Not
$ perl -e'm/foo/g' cheers, -- |
From @demerphqOn 5 February 2012 22:33, Ronald J Kimball <rjk@tamias.net> wrote:
I stand corrected, thanks for pointing that out, I definitely did not know that. $ perl -wle'sub foobar{"zoo"}; sub zoo { "giraffe" }; $_=qq("foo"); I do wonder if this is intentional or an interesting implementation side effect. Yves -- |
From tchrist@perl.com
It's been that way since time immemorial, but I can no longer --tom |
From @pjcjOn Mon, Feb 06, 2012 at 04:18:38AM -0700, Tom Christiansen wrote:
And it's seen far less frequently nowadays. Back then we had a general -- |
From tchrist@perl.comPaul Johnson <paul@pjcj.net> wrote
What, /ee? It's the normal way to get single variables interpolated. A single /e is like an eval { }, and compiled at compile time. But I never much go beyond that, because that's writing --tom |
From @ikegamiOn Mon, Feb 6, 2012 at 5:42 AM, demerphq <demerphq@gmail.com> wrote:
You said "Huh?!" when I said g, c and p are match and substitution operator Perl gives errors when you use "g" and "c" as regex flags. $ perl -e'qr/foo/c' $ perl -e'qr/foo/g' $ perl Useless (?g) - use /g modifier in regex; marked by <-- HERE in m/(?g <-- $ perl Useless (?c) - use /gc modifier in regex; marked by <-- HERE in m/(?c <-- $ perl -we'use re "/g";' $ perl -we'use re Unknown regular expression flag "c" at -e line 1 [It doesn't give errors for /p, though. but I don't see how that makes any which means "the m// modifier". Not
Exactly. Since qr// doesn't actually match, only regex pattern flags can be - Eric |
From @demerphqOn 6 February 2012 20:48, Eric Brine <ikegami@adaelis.com> wrote:
Ah, I see, this is a misunderstanding. "match and substitution
Here you seem to use "regex flags" where I would say "pattern
Do you find it productive when someone says something that you did not
This one is pretty obvious. It does not make any sense for _part_ of a
ditto.
I guess somebody MIGHT want every match to be global, but it seems
Ditto.
Tom explained this already. /p can be embedded in any pattern and it Having said that, in hindsight it probably was a mistake to allow /p on a qr//.
I find the term "regex flags" to be ambiguous and "Regex pattern 1. Pattern semantics flags: /m /i /s /u /d /l /a These modifiers change the meaning of specific meta-character, or 2. Pattern compilation flags: /x /o These patterns change the process by which a pattern is compiled. /x 3. Match control flags: /g /c These control how the matching is to proceed, and only apply to m// and s///. 4. Substitutions flags: /e /r These control the behavior of the substitution operator, and apply only to s/// Unfortunately this leaves /p out, as it does not fit well into any of Yves -- |
From @ikegamiOn Mon, Feb 6, 2012 at 9:23 PM, demerphq <demerphq@gmail.com> wrote:
oops, typo!
Here you seem to use "regex flags" where I would say "pattern
Noted. And yes of course Perl complains when you use a modifier which controls how the match is performed on a
That's good. That's what I would expect. The problem is that perlre lumps I find the term "regex flags" to be ambiguous and "Regex pattern
perlre simply calls them "modifiers" and "flags modifiers". The re pragma
Sounds good. A few comment/nits: I'm not sure a distinction is needed between (1) and (2), and it will lead perlop lists /ee in addition to /e, so /ee should be listed explicitly here. Because of where it can be used, I'd place /p under (2). (I believe it - Eric |
From @b2gillsOn Mon, Feb 6, 2012 at 11:04 PM, Eric Brine <ikegami@adaelis.com> wrote:
...
While (1) and (2) could be grouped together by what they can apply to; The phrase: If you wanted to describe both (1) and (2): I would like to note that the important words from (1) and (2) Also there is actually a difference between (1) and (2), ( Perhaps we should refer to qr// as a pattern. Just a random thought. )
If we do refer to /ee, perhaps we should note that you can append as
I agree with this. |
From @cpansproutOn Mon Feb 06 03:49:30 2012, tom christiansen wrote:
A single /e just makes the rhs an expression with no ‘eval’ and no block.
And the number of evals is one less than the number of e’s. -- Father Chrysostomos |
From tchrist@perl.com"Father Chrysostomos via RT" <perlbug-followup@perl.org>
I was using eval in the sense of "to evaluate an expression", not eval in 0. s/PATTERN/QQ_STRING/; In case 1, the replacement value is the result of evaluating EXPR. I don't think it is forbidden to speak of evaluating expressions
The number of evaluations is equal to the number of e’s. --tom |
From @cpansproutOn Sun Feb 12 07:18:57 2012, tom christiansen wrote:
I was wrong about there being no block. There is a block in s///e, but
‘Evaluate’ is fine, but don’t call it ‘eval’, as that refers to Perl’s
-- Father Chrysostomos |
From @cpansproutOn Sun Feb 05 09:55:52 2012, demerphq wrote:
The re pragma has always warned about invalid flags, rather than $ perl5.6.2 -e 'use re "/a"' -- Father Chrysostomos |
From @cpansproutOn Sun Feb 05 10:28:38 2012, tom christiansen wrote:
qr// is allowed on the rhs of =~ simply because arbitrary expressions
It’s useful for working around bugs. $u = ",echle etn sJ"; That example uses substitution, so the workaround is my $qr = qr/(.)/ -- Father Chrysostomos |
From @jkeenanOn Sun Feb 12 11:33:28 2012, sprout wrote:
Discussion in this RT petered out about 12 months ago. Is there still a bug in Perl which needs addressing? Thank you very much. |
Migrated from rt.perl.org#109798 (status was 'open')
Searchable as RT109798$
The text was updated successfully, but these errors were encountered: