🦋 Welcome to the IRC channel of the core developers of the Raku Programming Language (raku.org #rakulang). This channel is logged for the purpose of history keeping about its development | evalbot usage: 'm: say 3;' or /msg camelia m: ... | Logs available at irclogs.raku.org/raku-dev/live.html | For MoarVM see #moarvm
Set by lizmat on 8 June 2022.
Geth rakudo/main: f4c228dcf3 | (Elizabeth Mattijsen)++ | src/Raku/Actions.nqp
RakuAST: DRY the codegen of simple variables
 09:08
[Tux] Rakudo v2023.08-200-gf4c228dcf (v6.d) on MoarVM 2023.08-3-g2f8234c22
csv-ip5xs1.003 - 1.103
csv-ip5xs-205.326 - 5.999
csv-parser4.004 - 4.478
csv-test-xs-200.345 - 0.347
test6.379 - 7.725
test-t1.506 - 1.645
test-t --race0.899 - 1.087
test-t-2021.044 - 21.547
test-t-20 --race6.415 - 7.314
 10:05
Geth rakudo/main: f67797dbe2 | (Elizabeth Mattijsen)++ | src/Raku/Actions.nqp
RakuAST: only join if necessary
 10:10
finanalyst .tell @coleman the docs-website buildkite agent is delaying a branch build. Not sure why 10:14
tellable6 finanalyst, I cannot recognize this command. See wiki for some examples: github.com/Raku/whateverable/wiki/Tellable
finanalyst .tell coleman: the docs-website buildkite agent is delaying a branch build. Not sure why 10:15
tellable6 finanalyst, I'll pass your message to coleman
Geth rakudo/main: eb8fea816f | (Elizabeth Mattijsen)++ | src/Raku/Grammar.nqp
RakuAST: streamline special-variable tokens

This is mostly about not needing to capture <sym> for the error message.		 10:24
Geth rakudo/main: 9d711ea512 | (Elizabeth Mattijsen)++ | src/Raku/Grammar.nqp
RakuAST: remove some more unneeded captures
 10:43
rakudo/main: 863373a6c1 | (Elizabeth Mattijsen)++ | 2 files
RakuAST: give .intern-int a default radix of 10

For better readability, and smaller bytecode at the caller's side		 10:59
Geth rakudo/main: 3a91b4956e | (Elizabeth Mattijsen)++ | src/Raku/Actions.nqp
RakuAST: streamline parameter / arglist actions

  - prevent repeated hash lookups where possible
  - prevent duplicate condition evaluation
  - class lookup once if needed multiple times		 12:16
coleman Looking into it 14:59
tellable6 2023-09-12T10:15:02Z #raku-dev <finanalyst> coleman: the docs-website buildkite agent is delaying a branch build. Not sure why
ab5tract lizmat: WRT the remaining once block issue.. 15:44
ab5tract I've determined that the var declaration is getting generated on the wrong scope.. there is a QAST::Block definition in between the declaration and the check on p6stateassign, meaning that this op is never true because there is no statevar in the current block 15:46
lizmat hmmm
ab5tract or at least, that's my working assumption at the moment 15:47
*p6stateinit 15:49
the mystery also deepens:
m: 'my int $i; $i += ($ = 21) for 1,2; say $i'
camelia WARNINGS for <tmp>:
Useless use of constant string "my int $i; $i += ($ = 21) for 1,2; say $i" in sink context (line 1)
ab5tract m: Q| my int $i; $i += ($ = 21) for 1,2; say $i | .AST.EVAL 15:50
camelia 42
ab5tract m: 'my int $i; $i += (state $s = 21) for 1,2; say $i'
camelia WARNINGS for <tmp>:
Useless use of constant string "my int $i; $i += (state $s = 21) for 1,2; say $i" in sink context (line 1)
ab5tract doh! 15:51
m: my int $i; $i += ($ = 21) for 1,2; say $i
camelia 42
ab5tract m: Q| my int $i; $i += ($ = 21) for 1,2; say $i | .AST.EVAL
camelia 42
ab5tract m: my int $i; $i += (state $s = 21) for 1,2; say $i
camelia Use of uninitialized value of type Any in numeric context
0
in block at <tmp> line 1
Use of uninitialized value of type Any in numeric context
in block at <tmp> line 1
ab5tract m: Q| my int $i; $i += (state $s = 21) for 1,2; say $i | .AST.EVAL 15:52
camelia Use of uninitialized value of type Any in numeric context
0
in block <unit> at /home/camelia/EVAL_0 line 1
Use of uninitialized value of type Any in numeric context
in block <unit> at /home/camelia/EVAL_0 line 1
ab5tract this implies that p6stateassign isn't true when the state variable is defined directly on the Block named 'unit' 15:54
gar, *p6stateinit
lizmat ack 15:55
ab5tract but p6stateinit leads me into moar territory, IIUC
MasterDuke m: Q| say |.AST.EVAL 16:13
camelia ===SORRY!=== Error while compiling
Unsupported use of bare "say". In Raku please use: .say if to call it
as a method on $_, or use an explicit invocant or argument, or use &say
to refer to the function as a noun.
------> say⏏ …
Geth rakudo/main: 636c878946 | (Elizabeth Mattijsen)++ | src/Raku/Actions.nqp
RakuAST: some massive streamlining

  - more shortcuts
  - multiple self.attach: $/, … combined into a single with conditional
   arguments to reduce bytecode size
  - generally combining stuff into ternaries
  - using postfix if where it makes sense
Takes 1 second off spectest. Less than I hoped :-(		 17:52
lizmat still, about .5% 18:00
MasterDuke lizmat: any idea why some *Method Nodes take an ast and some just a name? e.g., here github.com/rakudo/rakudo/commit/63...1158-R1166 18:09
lizmat if the name can only be \w+ basically, it is a string. Otherwise it's a RakuAST::Name object 18:10
MasterDuke ah 18:12
lizmat must admit this is potentially something that may need to be made consistent
or at least rename RakuAST::Name to FQName or something 18:13
MasterDuke yep 18:14
`if $foo -> $bar { ... }` is (still) slower than `if $foo { my $bar = $foo; ... }`, right? 18:15
lizmat well, yeah, but that appears to be because that will prevent the outer scope from being inlined 18:16
MasterDuke and is that what inspired your comment over in #moarvm ?
lizmat yes
MasterDuke hm
lizmat raku -MSIL -e 'sub a($a) { if $a -> $b { $b } }; a(42) for ^3000000' 18:17
MasterDuke are you trying to remove all duplicated uses of `$<foo>`?
lizmat vs: 18:18
raku -MSIL -e sub a($a) { if my $b := $a { $b } }; a(42) for ^3000000
yeah, as those are hash lookups, and they're relatively expensive, no?
MasterDuke yeah, i was just looking at that most recent diff 18:19
so github.com/rakudo/rakudo/commit/63...a2aa0R1903 could make sense to pull out of the `if`, right?
lizmat right, except that that slows down 18:20
m: sub a($a) { if $a -> $b { $b } }; a(42) for ^300000; say now - INIT now
camelia 0.060584146
lizmat m: sub a($a) { if my $b := $a { $b } }; a(42) for ^3000000; say now - INIT now
camelia 0.100177813
lizmat that's weird
MasterDuke well, i meant just move the line above the if
lizmat m: sub a($a) { if $a -> $b { $b } }; a(42) for ^300000; say now - INIT now 18:21
camelia 0.061284621
lizmat m: sub a($a) { if my $b := $a { $b } }; a(42) for ^3000000; say now - INIT now
camelia 0.097742603
MasterDuke i.e., `my $desigilname := $<desigilname>; if $desigilname {`
the := version takes slightly under half the time of the -> version locally
lizmat yeah, for me as well 18:22
on an M1
MasterDuke github.com/rakudo/rakudo/commit/63...a2aa0R1950 and github.com/rakudo/rakudo/commit/63...a2aa0R1989 are other places with `if $<foo> { my $foo = $<foo>`. but maybe polluting the outer scope 18:23
isn't worth removing only one more $<foo>
bunch more of those 18:24
lizmat well.. I think the inline stopping is wrong, unless you can tell me otherwise :-)
why would sub a($a) { if $a -> $b { } } stop inlining of sub a ? 18:25
dinner& 18:26
MasterDuke m: my $a = 42; my $b = +(rand > 0.5); if $b -> $a { say $a }; say $a 18:28
camelia 1
42
MasterDuke not sure 18:29
nemokosch are declarations meant to work with statement modifiers? 20:20
or maybe the more useful question would be: which declarations are meant to work with statement modifiers? 20:21
MasterDuke lizmat: what do you mean by your commit message for github.com/rakudo/rakudo/commit/c1...fa7d794ffb ? 20:56
when you say this stuff shouldn't be copied over to the raku grammar, is that because you think it doesn't belong in the *grammar* (and should be implemented elsewhere)? or should just be removed completely? 20:58
vrurg m: say 20:59
camelia ===SORRY!===
Argument to "say" seems to be malformed
at <tmp>:1
------> say⏏<EOL>
Other potential difficulties:
Unsupported use of bare "say". In Raku please use: .say if you meant
to call it as a method on $_, or u…
vrurg m: put
camelia ===SORRY!===
Argument to "put" seems to be malformed
at <tmp>:1
------> put⏏<EOL>
Other potential difficulties:
Function "put" may not be called without arguments (please use () or whitespace to denote arguments, or &put t…
vrurg m: note
camelia Noted
vrurg I'd be for removing it altogether. It's like warning about a user/imported routine invoked without arguments. Besides, it's inconsistent. 21:00
I guess it was introduced back in days when migration from Perl was expected. For now it just makes little to no sense. 21:01
MasterDuke that'll have to be 6.e change 21:02
vrurg Sure. 21:04
MasterDuke but that does mean we have to implement it for rakuast first 21:06
nemokosch I don't know. It seems kinda weird to consider these obsolete errors a part of the language, rather than just additional courtesy from a parser 21:09
vrurg Waste of resources... :(
nemokosch just think about it. Say, the migration from 5 to 6 had happened, in an alternate reality. How long would these notifiers have lasted? 21:10
Arguably, shorter than the code written for the new version 21:11
lizmat MasterDuke: I think you got your answer, right ? 22:13
MasterDuke can we do language specific stuff in the grammar? 22:17
*language version 22:18
lizmat yes, e.g. to handle prefix // 22:41
look for nqp::getcomp('Raku').language_revision 22:42
sleep&
MasterDuke thanks

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!