|
09:11
dakkar joined
14:07
hudo_ left
14:08
hudo joined,
hudo_ joined
|
|||
| rcmlz | Not all what you read e.g. about Java OO and Best Practice/Design Pattern is relevant for Raku, as Raku has advanced features like multi methods that Java has not. So you can solve a problem maybe easier using Rakus features than you would in Java. | 14:51 | |
| So the standard educational task „Given a rather complex (theoretical) class hierachy with lots of inheritance and overwriting, which method gets actually called given a certain instance and certain arguments“ is maybe easier to solve 🤓 | 15:04 | ||
| antononcube | I should make for Raku something like this GoFBot. | 15:07 | |
| rcmlz | I recently took a course at university about OO that demonstrated various OO Concepts and Pitfalls (and proved that Java OO is not sound, if I remember right) using various Languages (mainly Java, C++, Rust) - sometimes I tried the examples/task in Raku and was often surprised that it actually worked. | 15:22 | |
| If it is possible to generate high quality Raku, working code for/from a OO Design Pattern List from a on that matter Textbook - would be interesting. | 15:27 | ||
| Sort of for all patterns -> pattern: say „Pattern: Java Solution vs. Raku Solution“ | 15:30 | ||
| 🤓 | |||
| antononcube | Well, it does not seem to be that hard. Again, testing and testers are needed. My experience is that that GoFBot produced correct code 80% of the time. | 15:33 | |
| BTW, GoFBot uses UML in Mermaid-JS. | 15:34 | ||
| It is related to my TODO list of "UML::Translators" -- to translate UML code into C++, Java, Raku, etc. raku.land/zef:antononcube/UML::Translators | 15:35 | ||
| rcmlz | Interesting, did not know that „Raku-Code to PlantUML/Mermaid“ exists. | 15:41 | |
|
16:15
Manifest0 joined
|
|||
| Manifest0 | Hi! What's the difference between invoking a function this way function(1) or function:1 ? | 16:17 | |
| in my pc: | 16:18 | ||
| [8] > my @a; my $t = time; for ^10000000 {@a.push($_)}; say "Took: {time - $t} seconds" | |||
| Took: 23 seconds | |||
| [8] > my @a; my $t = time; for ^10000000 {@a.push:$_}; say "Took: {time - $t} seconds" | |||
| Took: 9 seconds | |||
| that's a big difference | |||
| lizmat | Manifest0: which version of Rakudo are you using? | 16:20 | |
| fwiw, I don't see a difference in performance between the two syntaxes | |||
| Manifest0 | Welcome to Rakudo™ v2024.12-33-ge7f9223c5. | ||
| lizmat | m: my @a; my $t = time; for ^1000000 {@a.push($_)}; say "Took: {time - $t}" | ||
| camelia | Took: 1 | ||
| lizmat | m: my @a; my $t = time; for ^1000000 {@a.push: $_}; say "Took: {time - $t}" | ||
| camelia | Took: 1 | ||
| lizmat | fwiw, it's better to use "now" for these cases | 16:21 | |
| m: my @a; my $t = now; for ^1000000 {@a.push: $_}; say "Took: {now - $t}" | |||
| camelia | Took: 0.543170965 | ||
| lizmat | m: my @a; my $t = now; for ^1000000 {@a.push($_)}; say "Took: {now - $t}" | ||
| camelia | Took: 0.510747792 | ||
| lizmat | m: my @a; my $t = now; for ^1000000 {@a.push($_)}; say "Took: {now - $t}" | ||
| camelia | Took: 0.52691798 | ||
| lizmat | m: my @a; my $t = now; for ^1000000 {@a.push: $_}; say "Took: {now - $t}" | ||
| camelia | Took: 0.572151738 | ||
| Manifest0 | let me try | 16:22 | |
| m: my @a; my $t = time; for ^10000000 {@a.push($_)}; say "Took: {now - $t}" | |||
| camelia | Took: Instant:46.396049891 | ||
| Manifest0 | m: my @a; my $t = time; for ^10000000 {@a.push:$_}; say "Took: {now - $t}" | ||
| lizmat | both cases: s/time/now :-) | ||
| camelia | Took: Instant:41.327424937 | ||
| Manifest0 | got it | 16:23 | |
| m: my @a; my $t = now; for ^10000000 {@a.push:$_}; say "Took: {now - $t}" | |||
| camelia | Took: 3.792938414 | ||
| Manifest0 | m: my @a; my $t = now; for ^10000000 {@a.push($_)}; say "Took: {now - $t}" | ||
| camelia | Took: 8.27316351 | ||
| Manifest0 | already some difference | 16:24 | |
| lizmat | hmmm that's weird | ||
| are you on Intel hardware ? | |||
| Manifest0 | yes | ||
| lizmat | ok, then that's maybe the reason I can't repro as I'm on Apple silicon | ||
| and that doesn't have the JIT | 16:25 | ||
| m: my @a; my $t = now; for ^10000000 {@a.push($_)}; say "Took: {now - $t}" | |||
| camelia | Took: 8.101594819 | ||
| lizmat | m: my @a; my $t = now; for ^10000000 {@a.push: $_}; say "Took: {now - $t}" | ||
| camelia | Took: 8.216274028 | ||
| lizmat | I wonder whether the CPU of camelia is now being throttled, but weird it is | 16:26 | |
| m: my @a; my $t = now; for ^10000000 {@a.push: $_}; say "Took: {now - $t}" | |||
| camelia | Took: 8.28036817 | ||
| lizmat | m: my @a; my $t = now; for ^10000000 {@a.push($_)}; say "Took: {now - $t}" | ||
| camelia | Took: 8.714330314 | ||
| lizmat | would you mind making a Rakudo issue for this? | 16:27 | |
| Manifest0 | sure | ||
| lizmat | also: could you try locally running these tests with MVM_JIT_DISABLE=1 ? | 16:28 | |
| and see if that makes the results more consistent ? | |||
| Manifest0 | sure | 16:29 | |
| damn, i'm not even now able to reproduce it on my machine.... But i ran several times and with parens was hitting the 23 seconds mark, while with ':' it was in the 9 secs | 16:32 | ||
| should i still open a github issue? I'm not able to reproduce it anymore :-( | 16:37 | ||
| lizmat | meh | ||
| well, I guess we need to wait until we can repro | 16:38 | ||
| rcmlz | On glot.io/snippets/h3gee35qz6 : is faster than . | 16:48 | |
|
16:51
MasterDuke joined
|
|||
| Manifest0 | i can reproduce it in the REPL. I copied all the steps i've made before (errors included), and i have the same results | 16:51 | |
| MasterDuke | `push:$_` has different timing (and ast) than `push: $_` (note the space after the colon) | 16:52 | |
| lizmat | there shouldn't be any difference in the QAST produced, is there ? | 16:53 | |
| Manifest0 | paste.opensuse.org/pastes/44988037d045 | 16:54 | |
| MasterDuke | lizmat: there is | ||
| lizmat | weird | ||
| Manifest0 | yes, that's the space! | 16:56 | |
| with the space it takes the same amount of time | |||
| without the space is much faster | |||
| MasterDuke | and a difference persists at --target=optimize also (somewhat obviously since we see a time difference) | ||
| but without the space it doesn't work. i.e., the $_ values aren't in the array | 16:57 | ||
| Manifest0 | shouldn't we get a warning then? | 17:00 | |
| MasterDuke | i'd say yes | 17:08 | |
| afk now | 17:09 | ||
|
17:14
MasterDuke left
|
|||
| lizmat | ok, found the reason for the discrepancy | 17:36 | |
| @a.push: $_ is codegenned as @a.push($_), as one would expect | |||
|
17:37
dakkar left
|
|||
| lizmat | @a.push:$_ on the other hand is codegenned as @a.push( :$_), aka with the named argument "_" taking the value of $_ | 17:37 | |
| and the named argument is silently ignored | 17:38 | ||
| effectively making it: | |||
| @a.push | |||
| this would have been caught sooner if we would have a way to mark method calls as "endpoints" in the sense that they don't accept any unknown named arguments | 17:41 | ||
| Manifest0 | lizmat: nice finding! | 18:21 | |
| lizmat | yeah, looking at the RakuAST tree made it clear immediately | ||
| m: say Q|my @a; @a.push:$_|.AST.statements[1].expression | 18:22 | ||
| camelia | RakuAST::ApplyPostfix.new( operand => RakuAST::Var::Lexical.new("\@a"), postfix => RakuAST::Call::Method.new( name => RakuAST::Name.from-identifier("push"), args => RakuAST::ArgList.new( RakuAST::ColonPair::Variable.new( … |
||
| lizmat | the "RakuAST::ColonPair::Variable" gives that away | 18:23 | |
| m: say Q|my @a; @a.push:$_|.AST.statements[1].expression.postfix | |||
| camelia | RakuAST::Call::Method.new( name => RakuAST::Name.from-identifier("push"), args => RakuAST::ArgList.new( RakuAST::ColonPair::Variable.new( key => "_", value => RakuAST::Var::Lexical.new("\$_") ) ) ) |
||
| lizmat | RakuAST::ColonPair::Variable.new( key => "_", value => RakuAST::Var::Lexical.new("\$_") | ||
|
22:34
kjp left
22:36
kjp joined
22:38
kjp left,
kjp joined
|
|||