Reimplement List.min in Perl6 instead of PIR

[p6rt]

Migrated from rt.perl.org#63712 (status was 'resolved')

Searchable as RT63712$

[p6rt]

From @bacek

---

src/builtins/any-list.pir | 47 --------------------------------------------- src/setting/Any-list.pm | 20 ++++++++++++++++++- 2 files changed, 19 insertions(+), 48 deletions(-)

[p6rt]

From @bacek

059ec2c89dc266165efcee344f248218be567fe4.diff

```diff diff --git a/src/builtins/any-list.pir b/src/builtins/any-list.pir index b87666b..bc5cc9e 100644 --- a/src/builtins/any-list.pir +++ b/src/builtins/any-list.pir @@ -223,53 +223,6 @@ Return a List with the keys of the invocant. .return(res) .end -=item min - -=cut - -.namespace [] -.sub 'min' :multi() - .param pmc values :slurpy - .local pmc by - by = get_hll_global 'infix:cmp' - unless values goto have_by - $P0 = values[0] - $I0 = isa $P0, 'Sub' - unless $I0 goto have_by - by = shift values - have_by: - .tailcall values.'min'(by) -.end - - -.namespace ['Any'] -.sub 'min' :method :multi(_) - .param pmc by :optional - .param int has_by :opt_flag - if has_by goto have_by - by = get_hll_global 'infix:cmp' - have_by: - - .local pmc it, result - $P0 = self.'list'() - it = $P0.'iterator'() - unless it goto fail - result = shift it - loop: - unless it goto done - $P0 = shift it - $I0 = by($P0, result) - unless $I0 < 0 goto loop - result = $P0 - goto loop - fail: - .local num failres - failres = "+Inf" - .return (failres) - done: - .return (result) -.end - .namespace [] .sub 'max' :multi() diff --git a/src/setting/Any-list.pm b/src/setting/Any-list.pm index 5758dce..0aaeccc 100644 --- a/src/setting/Any-list.pm +++ b/src/setting/Any-list.pm @@ -23,7 +23,20 @@ class Any is also { $res = &$expression($res, |@args); } $res; - } + }; + + # RT #63700 - parse failed on &infix: + our Array multi method min( $values: Code $by = sub { $^a cmp $^b } ) { + my @list = $values.list; + return +Inf unless @list.elems; + my $res = @list.shift; + for @list -> $x { + if (&$by($res, $x) > 0) { + $res = $x; + } + } + $res; + }; } our List multi grep(Code $test, *@values) { @@ -34,4 +47,9 @@ multi reduce ( Code $expression ;; *@values ) { @values.reduce($expression); } +our List multi min(*@values) { + my $by = @values[0] ~~ Code ?? shift @values !! sub { $^a cmp $^b }; + @values.min($by); +} + # vim: ft=perl6 ```

[p6rt]

From @moritz

On Sat Mar 07 21​:50​:00 2009, bacek wrote​:

--- src/builtins/any-list.pir | 47 --------------------------------------------- src/setting/Any-list.pm | 20 ++++++++++++++++++- 2 files changed, 19 insertions(+), 48 deletions(-)

Applied as a7214ac28c5c7c47932f1e76a15c8707524f964d, thanks.

Moritz

[p6rt]

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

[p6rt]

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

[p6rt]

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

[p6rt]

From @pmichaud

On Sat, Mar 07, 2009 at 09​:50​:02PM -0800, Vasily Chekalkin wrote​:

+our List multi min(*@​values) { + my $by = @​values[0] ~~ Code ?? shift @​values !! sub { $^a cmp $^b }; + @​values.min($by); +}

This doesn't match the spec -- the $by parameter is required. At any rate, the first argument should not be explicitly checked for being a Code object -- if we do this it should be via a multi signature.

Yes, the PIR code was "cheating" in this respect, but I don't want to blindly copy our cheats from PIR into the setting without subjecting them to another review.

+ # RT #​63700 - parse failed on &infix​:\ + our Array multi method min( $values​: Code $by = sub { $^a cmp $^b } ) { + my @​list = $values.list; + return +Inf unless @​list.elems; + my $res = @​list.shift; + for @​list -> $x { + if (&$by($res, $x) > 0) { + $res = $x; + } + } + $res; + }; }

Why are C\ and C\ specced as returning C\ and C\? Shouldn't they just return the single element that is the minimum or maximum from the list?

Pm

[p6rt]

From @moritz

Patrick R. Michaud wrote​:

On Sat, Mar 07, 2009 at 09​:50​:02PM -0800, Vasily Chekalkin wrote​:

+our List multi min(*@​values) { + my $by = @​values[0] ~~ Code ?? shift @​values !! sub { $^a cmp $^b }; + @​values.min($by); +}

This doesn't match the spec -- the $by parameter is required. At any rate, the first argument should not be explicitly checked for being a Code object -- if we do this it should be via a multi signature.

I've fixed that, and corrected the spec tests accordingly.

Yes, the PIR code was "cheating" in this respect, but I don't want to blindly copy our cheats from PIR into the setting without subjecting them to another review.

+ # RT #​63700 - parse failed on &infix​:\ + our Array multi method min( $values​: Code $by = sub { $^a cmp $^b } ) { + my @​list = $values.list; + return +Inf unless @​list.elems; + my $res = @​list.shift; + for @​list -> $x { + if (&$by($res, $x) > 0) { + $res = $x; + } + } + $res; + }; }

Why are C\ and C\ specced as returning C\ and C\? Shouldn't they just return the single element that is the minimum or maximum from the list?

I agree, so I removed the non-sense return types from the specs.

Cheers, Moritz

[p6rt]

From @bacek

Patrick R. Michaud via RT wrote​:

On Sat, Mar 07, 2009 at 09​:50​:02PM -0800, Vasily Chekalkin wrote​:

+our List multi min(*@​values) { + my $by = @​values[0] ~~ Code ?? shift @​values !! sub { $^a cmp $^b }; + @​values.min($by); +}

This doesn't match the spec -- the $by parameter is required. At any rate, the first argument should not be explicitly checked for being a Code object -- if we do this it should be via a multi signature.

Yes, the PIR code was "cheating" in this respect, but I don't want to blindly copy our cheats from PIR into the setting without subjecting them to another review.

This is workaround for this situation​:

\ rakudo​: multi sub foo(Code &x, *@​values) {...}; multi sub foo(*@​values) {...}; say foo(sub {}, 1,2,3); \ rakudo 2daf6b​: OUTPUT«Ambiguous dispatch to multi 'foo'. Ambiguous candidates had signatures​:␤​:(Code x, Any @​values)␤​:(Any @​values)␤␤current instr.​: '_block14' pc 101 (EVAL_18​:52)␤»

Should I fill bug report for this?

-- Bacek

[p6rt]

From @pmichaud

On Mon Mar 09 00​:41​:24 2009, bacek wrote​:

Patrick R. Michaud via RT wrote​:

At any rate, the first argument should not be explicitly checked for being a Code object -- if we do this it should be via a multi signature.

Yes, the PIR code was "cheating" in this respect, but I don't want to blindly copy our cheats from PIR into the setting without subjecting them to another review.

This is workaround for this situation​:

\ rakudo​: multi sub foo(Code &x, *@​values) {...}; multi sub foo(*@​values) {...}; say foo(sub {}, 1,2,3); \ rakudo 2daf6b​: OUTPUT«Ambiguous dispatch to multi 'foo'. Ambiguous candidates had signatures​:␤​:(Code x, Any @​values)␤​:(Any @​values)␤␤current instr.​: '_block14' pc 101 (EVAL_18​:52)␤»

Should I fill bug report for this?

Yes please. Closing this ticket, thanks!

Pm

[p6rt]

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