search

Raku / old-issue-tracker

Tickets from RT
https://github.com/Raku/old-issue-tracker/issues
2 stars 1 forks source link

[]{} method2sub broke *[0]([1, 2, 3]) #3241

Closed p6rt closed 10 years ago

p6rt commented 11 years ago

Migrated from rt.perl.org#120025 (status was 'resolved')

Searchable as RT120025$

p6rt commented 11 years ago

From @lizmat

TimToady pointed out that my work on converting [] and {} accesses to sub calls, rather than method calls, had one more casualty​:

*[0]([1, 2, 3])

This is tested in t/spec/S02-types/whatever.t, line 252 and currently fudged.

The suggested change in whatever_curry did not give the right result. And now I'm in here over my head, so I'm submitting a bug report as promised.

Liz

[18​:53​:51] \ lizmat​: did you notice that the subscripting change broke *.[], *.\<>, and *.{} [18​:54​:21] \ TimToady​: no, don't think so [18​:54​:55] \ I fudged a number of tests, but I don't recall fudging any .[] related tests [18​:54​:56] \ nr​: my @​a = [1,2,3],[4,5,6],[7,8,9]; say @​a.map​: *.[2] [18​:54​:59] \ niecza v24-95-ga6d4c5f​: OUTPUT«3 6 9␤» [18​:55​:00] \ ..rakudo 50a57d​: OUTPUT«Index out of range. Is​: 2, should be in 0..0␤ in method \ at src/gen/CORE.setting​:11768␤ in any at src/gen/Metamodel.nqp​:2671␤ in any find_method_fallback at src/gen/Metamodel.nqp​:2659␤ in any find_method at src/gen/Metamodel.nqp​:946␤ in method reify…» [19​:09​:11] \ jnthn​: I'm not sure how to fix .[] etc., any pointers ? [19​:11​:03] \ aaahhhhh… I completely missed that TimToady meant Whatever.[] [19​:11​:16] \ yes, that broke​: simplest case [19​:11​:37] \ nr say *[0](1,2,3) [19​:11​:42] \ nr​: say *[0](1,2,3) [19​:11​:45] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Excess arguments to ANON, used 1 of 3 positionals␤ at /tmp/GbWfbqMj6Y line 0 (ANON @​ 1) ␤ at /tmp/GbWfbqMj6Y line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE…» [19​:11​:45] \ ..rakudo 50a57d​: OUTPUT«No such method 'postcircumfix​:\<( )>' for invocant of type 'Whatever'␤ in block at /tmp/IcpYdTyvhm​:1␤␤» [19​:12​:04] \ lizmat​: yeah [19​:12​:10] \ lizmat​: See Actions.pm, whatever_curry [19​:12​:16] lizmat checks [19​:12​:31] \ lizmat​: Near the top it OKs infix, prefix and postfix. Try adding postcircumfix to that list. [19​:13​:36] \ I guess I should follow the callmethod value, right? [19​:14​:13] \ ? [19​:14​:25] \ no, inside whatever_curry [19​:14​:38] \ (nqp​::index($past.name, '&infix​:') == 0 || [19​:14​:38] \ nqp​::index($past.name, '&prefix​:') == 0 || [19​:14​:38] \ nqp​::index($past.name, '&postfix​:') == 0 || [19​:14​:42] \ ahhhh ok [19​:14​:42] \ After that [19​:16​:21] \ compiling and testing [19​:19​:26] \ using index to test for prefix bothers me, since it'll scan uselessly whenever it doesn't match [19​:20​:49] moritz waits for masak++ to construct a bug report that involves a variable named 'something&infix​:bla' which shouldn't be treated as an infix, but is [19​:21​:11] \ :P [19​:21​:21] \ well, the == 0 prevents that, but it's just the waste that bothers me [19​:24​:07] \ it's like using a pile-driver to squash a bug, and after doing it 50 times, then checking to see if you were successful the first time [19​:25​:33] \ might still be faster than nqp​::eq(nqp​::substr($past.name, 0, nqp​::chars('&infix​:')), '&infix​:') [19​:25​:38] \ and a smart index might even have set up Boyer-Moore tables on the assumption it was going to scan [19​:26​:18] \ sure would be nice if we could write Perl 6 in Perl 6 :) [19​:27​:11] \ then it's just /^ '&' [ in | pre | post 'circum'? ] fix '​:' / :) [19​:27​:58] \ $ perl6 --ll-exception -e 'say *[0](1,2,3)' [19​:27​:58] \ Too many positional parameters passed; got 3 but expected 1 [19​:28​:30] \ full stack trace​: https://gist.github.com/lizmat/6717575 [19​:28​:35] \ the error changed [19​:28​:54] \ but I have no idea where that code lives [19​:29​:40] \ lizmat​: maybe look at the ast [19​:30​:38] \ lizmat​: src/Perl6/Actions.nqp sez a recursive grep [19​:33​:00] \ sanity check​: *[0](1,2,3,4,5) should return the first element of the (1,2,3,4,5) parcel, right ? [19​:35​:02] \ nr​: say (1,2,3,4,5)[0] [19​:35​:05] \ rakudo 50a57d, niecza v24-95-ga6d4c5f​: OUTPUT«1␤» [19​:35​:08] \ so it would seem [19​:35​:23] \ right *phew* [19​:35​:35] \ so *() has to be exempted from the postcirumfixes for * [19​:36​:05] \ unless we force people to say (*[0])(1,2,3,45) [19​:36​:16] \ trying that [19​:36​:18] \ if we decide there's a use case for *.() [19​:36​:32] \ which arguably there probably is [19​:36​:55] \ map​: *.() should invoke each of a list of invokables [19​:37​:23] \ so I'm inclined to not make *() an exception, and force people to say (*[0])(1,2,3,4,5) if that's what they mean [19​:38​:06] \ in that case, *[0](1,2,3,4,5) means { $_.[0].(1,2,3,4,5) } (as a WhateverCode) [19​:38​:17] \ that feels cleaner to me [19​:38​:43] \ the use case for callinging a WhateverCode directly is minimal [19​:38​:50] \ *inginging [19​:39​:14] lizmat feels she no longer has to duck [19​:39​:36] \ nr​: say *[0](1,2,3,4,5).WHAT [19​:39​:36] \ as in​: this is way over my head [19​:39​:38] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Excess arguments to ANON, used 1 of 5 positionals␤ at /tmp/vU2WHEK3m3 line 0 (ANON @​ 1) ␤ at /tmp/vU2WHEK3m3 line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE…» [19​:39​:39] \ ..rakudo 50a57d​: OUTPUT«No such method 'postcircumfix​:\<( )>' for invocant of type 'Whatever'␤ in block at /tmp/uP5CQeMuln​:1␤␤» [19​:40​:06] \ n​: say (*[0](1,2,3,4,5)).WHAT [19​:40​:08] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Excess arguments to ANON, used 1 of 5 positionals␤ at /tmp/zSzO4DEzSC line 0 (ANON @​ 1) ␤ at /tmp/zSzO4DEzSC line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE…» [19​:40​:28] \ I guess niecza exempts .() though [19​:41​:07] \ n​: say (*(1,2,3,4,5)).WHAT [19​:41​:09] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Unable to resolve method postcircumfix​:\<( )> in type Whatever␤ at /tmp/JqJTGWQeaz line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4584 (module-C…» [19​:49​:02] \ n​: say ({.(1,2,3,4,5)}).WHAT [19​:49​:04] \ niecza v24-95-ga6d4c5f​: OUTPUT«(Block)␤» [19​:49​:23] TimToady thinks *(1,2,3,4,5) should mean that [19​:49​:39] \ well, except it's a WhateverCode [19​:51​:15] \ TimToady​: *(1,2,3,4,5) looks like -> &f { &f(1,2,3,4,5) } to me. [19​:51​:40] \ oh wait, we're saying the same thing, aren't we? [19​:51​:47] \ yes, though I just wrote it {.()} [19​:52​:36] \ so I flipflopped on my answer to lizmat++ and don't think .() should be an exception under Whatever [19​:53​:06] \ .oO( lalaalalalala I'm not listening ) :-) [19​:53​:49] \ but you BROKE it! WAAAH! :P [19​:54​:07] \ may I plead insanity ? [19​:54​:54] \ as long as you don't plead ultro-low-frequency EM [19​:55​:03] \ *ultra [19​:55​:17] \ My Hero! [19​:55​:55] \ but in that case someone needs to file a bug [19​:56​:16] \ .oO( I was suddenly reminded of http://en.wikipedia.org/wiki/My_Hero_(UK_TV_series) ) [19​:57​:08] TimToady looks around for an anti-hero to file the bug report for all the wrong reasons... [20​:00​:59] \ I will file one if I can't get it fixed

p6rt commented 10 years ago

From @Mouq

On Thu Sep 26 11​:17​:22 2013, elizabeth wrote​:

TimToady pointed out that my work on converting [] and {} accesses to sub calls, rather than method calls, had one more casualty​:

*[0]([1, 2, 3])

This is tested in t/spec/S02-types/whatever.t, line 252 and currently fudged.

Works now​:

$ perl6 -e'say *[0]([1, 2, 3])' 1

The suggested change in whatever_curry did not give the right result. And now I'm in here over my head, so I'm submitting a bug report as promised.

Liz

[18​:53​:51] \ lizmat​: did you notice that the subscripting change broke *.[], *.\<>, and *.{} [18​:54​:21] \ TimToady​: no, don't think so [18​:54​:55] \ I fudged a number of tests, but I don't recall fudging any .[] related tests [18​:54​:56] \ nr​: my @​a = [1,2,3],[4,5,6],[7,8,9]; say @​a.map​: *.[2] [18​:54​:59] \ niecza v24-95-ga6d4c5f​: OUTPUT«3 6 9␤» [18​:55​:00] \ ..rakudo 50a57d​: OUTPUT«Index out of range. Is​: 2, should be in 0..0␤ in method \ at src/gen/CORE.setting​:11768␤ in any at src/gen/Metamodel.nqp​:2671␤ in any find_method_fallback at src/gen/Metamodel.nqp​:2659␤ in any find_method at src/gen/Metamodel.nqp​:946␤ in method reify…» [19​:09​:11] \ jnthn​: I'm not sure how to fix .[] etc., any pointers ? [19​:11​:03] \ aaahhhhh… I completely missed that TimToady meant Whatever.[] [19​:11​:16] \ yes, that broke​: simplest case [19​:11​:37] \ nr say *[0](1,2,3) [19​:11​:42] \ nr​: say *[0](1,2,3) [19​:11​:45] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Excess arguments to ANON, used 1 of 3 positionals␤ at /tmp/GbWfbqMj6Y line 0 (ANON @​ 1) ␤ at /tmp/GbWfbqMj6Y line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE…» [19​:11​:45] \ ..rakudo 50a57d​: OUTPUT«No such method 'postcircumfix​:\<( )>' for invocant of type 'Whatever'␤ in block at /tmp/IcpYdTyvhm​:1␤␤» [19​:12​:04] \ lizmat​: yeah [19​:12​:10] \ lizmat​: See Actions.pm, whatever_curry [19​:12​:16] lizmat checks [19​:12​:31] \ lizmat​: Near the top it OKs infix, prefix and postfix. Try adding postcircumfix to that list. [19​:13​:36] \ I guess I should follow the callmethod value, right? [19​:14​:13] \ ? [19​:14​:25] \ no, inside whatever_curry [19​:14​:38] \ (nqp​::index($past.name, '&infix​:') == 0 || [19​:14​:38] \ nqp​::index($past.name, '&prefix​:') == 0 || [19​:14​:38] \ nqp​::index($past.name, '&postfix​:') == 0 || [19​:14​:42] \ ahhhh ok [19​:14​:42] \ After that [19​:16​:21] \ compiling and testing [19​:19​:26] \ using index to test for prefix bothers me, since it'll scan uselessly whenever it doesn't match [19​:20​:49] moritz waits for masak++ to construct a bug report that involves a variable named 'something&infix​:bla' which shouldn't be treated as an infix, but is [19​:21​:11] \ :P [19​:21​:21] \ well, the == 0 prevents that, but it's just the waste that bothers me [19​:24​:07] \ it's like using a pile-driver to squash a bug, and after doing it 50 times, then checking to see if you were successful the first time [19​:25​:33] \ might still be faster than nqp​::eq(nqp​::substr($past.name, 0, nqp​::chars('&infix​:')), '&infix​:') [19​:25​:38] \ and a smart index might even have set up Boyer-Moore tables on the assumption it was going to scan [19​:26​:18] \ sure would be nice if we could write Perl 6 in Perl 6 :) [19​:27​:11] \ then it's just /^ '&' [ in | pre | post 'circum'? ] fix '​:' / :) [19​:27​:58] \ $ perl6 --ll-exception -e 'say *[0](1,2,3)' [19​:27​:58] \ Too many positional parameters passed; got 3 but expected 1 [19​:28​:30] \ full stack trace​: https://gist.github.com/lizmat/6717575 [19​:28​:35] \ the error changed [19​:28​:54] \ but I have no idea where that code lives [19​:29​:40] \ lizmat​: maybe look at the ast [19​:30​:38] \ lizmat​: src/Perl6/Actions.nqp sez a recursive grep [19​:33​:00] \ sanity check​: *[0](1,2,3,4,5) should return the first element of the (1,2,3,4,5) parcel, right ? [19​:35​:02] \ nr​: say (1,2,3,4,5)[0] [19​:35​:05] \ rakudo 50a57d, niecza v24-95-ga6d4c5f​: OUTPUT«1␤» [19​:35​:08] \ so it would seem [19​:35​:23] \ right *phew* [19​:35​:35] \ so *() has to be exempted from the postcirumfixes for * [19​:36​:05] \ unless we force people to say (*[0])(1,2,3,45) [19​:36​:16] \ trying that [19​:36​:18] \ if we decide there's a use case for *.() [19​:36​:32] \ which arguably there probably is [19​:36​:55] \ map​: *.() should invoke each of a list of invokables [19​:37​:23] \ so I'm inclined to not make *() an exception, and force people to say (*[0])(1,2,3,4,5) if that's what they mean [19​:38​:06] \ in that case, *[0](1,2,3,4,5) means { $_.[0].(1,2,3,4,5) } (as a WhateverCode) [19​:38​:17] \ that feels cleaner to me [19​:38​:43] \ the use case for callinging a WhateverCode directly is minimal [19​:38​:50] \ *inginging [19​:39​:14] lizmat feels she no longer has to duck [19​:39​:36] \ nr​: say *[0](1,2,3,4,5).WHAT [19​:39​:36] \ as in​: this is way over my head [19​:39​:38] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Excess arguments to ANON, used 1 of 5 positionals␤ at /tmp/vU2WHEK3m3 line 0 (ANON @​ 1) ␤ at /tmp/vU2WHEK3m3 line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE…» [19​:39​:39] \ ..rakudo 50a57d​: OUTPUT«No such method 'postcircumfix​:\<( )>' for invocant of type 'Whatever'␤ in block at /tmp/uP5CQeMuln​:1␤␤» [19​:40​:06] \ n​: say (*[0](1,2,3,4,5)).WHAT [19​:40​:08] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Excess arguments to ANON, used 1 of 5 positionals␤ at /tmp/zSzO4DEzSC line 0 (ANON @​ 1) ␤ at /tmp/zSzO4DEzSC line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE…» [19​:40​:28] \ I guess niecza exempts .() though [19​:41​:07] \ n​: say (*(1,2,3,4,5)).WHAT [19​:41​:09] \ niecza v24-95-ga6d4c5f​: OUTPUT«Unhandled exception​: Unable to resolve method postcircumfix​:\<( )> in type Whatever␤ at /tmp/JqJTGWQeaz line 1 (mainline @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @​ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4584 (module-C…» [19​:49​:02] \ n​: say ({.(1,2,3,4,5)}).WHAT [19​:49​:04] \ niecza v24-95-ga6d4c5f​: OUTPUT«(Block)␤» [19​:49​:23] TimToady thinks *(1,2,3,4,5) should mean that [19​:49​:39] \ well, except it's a WhateverCode [19​:51​:15] \ TimToady​: *(1,2,3,4,5) looks like -> &f { &f(1,2,3,4,5) } to me. [19​:51​:40] \ oh wait, we're saying the same thing, aren't we? [19​:51​:47] \ yes, though I just wrote it {.()} [19​:52​:36] \ so I flipflopped on my answer to lizmat++ and don't think .() should be an exception under Whatever [19​:53​:06] \ .oO( lalaalalalala I'm not listening ) :-) [19​:53​:49] \ but you BROKE it! WAAAH! :P [19​:54​:07] \ may I plead insanity ? [19​:54​:54] \ as long as you don't plead ultro-low- frequency EM [19​:55​:03] \ *ultra [19​:55​:17] \ My Hero! [19​:55​:55] \ but in that case someone needs to file a bug [19​:56​:16] \ .oO( I was suddenly reminded of http://en.wikipedia.org/wiki/My_Hero_(UK_TV_series) ) [19​:57​:08] TimToady looks around for an anti-hero to file the bug report for all the wrong reasons... [20​:00​:59] \ I will file one if I can't get it fixed

It seems most of this conversation is about the behaviour of *(). I think a different report should be filed for *.() or the title of this report be changed. As an aside, what curries into a WhateverCode should probably be written in the spec somewhere…

p6rt commented 10 years ago

The RT System itself - Status changed from 'new' to 'open'

p6rt commented 10 years ago

From @usev6

On Sat Apr 12 19​:55​:27 2014, Mouq wrote​:

On Thu Sep 26 11​:17​:22 2013, elizabeth wrote​:

TimToady pointed out that my work on converting [] and {} accesses to sub calls, rather than method calls, had one more casualty​:

*[0]([1, 2, 3])

This is tested in t/spec/S02-types/whatever.t, line 252 and currently fudged.

Works now​:

$ perl6 -e'say *[0]([1, 2, 3])' 1

[...] It seems most of this conversation is about the behaviour of *(). I think a different report should be filed for *.() or the title of this report be changed. As an aside, what curries into a WhateverCode should probably be written in the spec somewhere…

Since said test runs fine and is no longer fudged, I'm closing this ticket.

p6rt commented 10 years ago

From @usev6

On Sat Apr 12 19​:55​:27 2014, Mouq wrote​:

On Thu Sep 26 11​:17​:22 2013, elizabeth wrote​:

TimToady pointed out that my work on converting [] and {} accesses to sub calls, rather than method calls, had one more casualty​:

*[0]([1, 2, 3])

This is tested in t/spec/S02-types/whatever.t, line 252 and currently fudged.

Works now​:

$ perl6 -e'say *[0]([1, 2, 3])' 1

[...] It seems most of this conversation is about the behaviour of *(). I think a different report should be filed for *.() or the title of this report be changed. As an aside, what curries into a WhateverCode should probably be written in the spec somewhere…

Since said test runs fine and is no longer fudged, I'm closing this ticket.

p6rt commented 10 years ago

@usev6 - Status changed from 'open' to 'resolved'