Parrot 4.7.0 "Hispaniolan" | parrot.org | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC
Set by moderator on 4 September 2012.
kid51 Boa noite, whiteknight 00:20
kid51 practices his Portuguese for class tomorrow
00:42 benabik joined
whiteknight hello kid51 00:48
02:09 MikeFair joined
MikeFair Good evening #parrot! :) 02:12
I am havinga bit of a problem coming up with the Grammar syntax that works for the "if statement" I'm trying to make 02:28
The main problem I'm having is that there's no clearly defined end point every time so I'm having difficulty figuring out how to create the rule for it 02:29
if <EXPR> then <statement>* [ else <statement>* ]? ['end if']? 02:30
Is kind of what I'm looking for
however if there is no 'else' or 'end' 'if' after the 'then' - only one single statement should be included in the <thenPart> 02:32
and if there is no 'end' 'if' after the 'else' -only one single statement should be included in the <elsePart> 02:33
Examples I've looked at have assumed only one 'block' or one 'statment' would follow the then or else because they have used {} or similar delimiters 02:34
I've considered trying to make 'then' .* ['else' | 'end' ] a "block" type thing 02:35
But it would have to be more like <?before 'else' | 'end'> 02:36
02:36 mvorl joined
MikeFair Thoughts/discussion on the issue would be welcomed - thx 02:37
hi there mvorl
mvorl Hi MikeFair & #parrot 02:38
sorear hello!
MikeFair hi there sorea! :) 02:39
err sorear even!
sorear MikeFair: if <EXPR> then <statement> [ <statement>* end if ]
extending this to handle else is left as an exercise.
MikeFair nods 02:40
sorear actually it should be if <EXPR> then [ <statement> [ <statement>* end if ]? || end if ]
MikeFair oh interesting 02:41
sorear this is an LL(infinity) grammar engine, you need to left-factor
MikeFair I had this was 'almost working'
ule thenPart {
'then' <statement>* <?before 'else' | 'end' 'if'>
| 'then' <statement>
but it didn't properly terminate when there was only one command and no end i 02:42
if
sorear your use of <?before> there seems quite wrong.
also, it's not left factored. 02:43
MikeFair Yeah I haven't been really mastering "left factoring" yet :)
It's gotten me into some cute trouble and I quickly learned "doctor it hurts when I infinitely recurse!" :) 02:44
MikeFair gives the above a shot.
the <?before> has seemingly been doing what I expect, "find the 'else' | 'end' 'if' and match the text before it 02:46
Explanations I've read online have explained it as like having a "run to this point" GOAL 02:47
but what I'm not clear on is how it will behave if the if statements are nested
sorear here's how it really works: 02:48
MikeFair I'm expecting the recursive part should justtake care of that
sorear regex position is before <statement>* <?before ...>
input looks like 1; 2; end if
MikeFair nodsl. 02:49
sorear first we try to match a statement because <statement>*
2; end if
now because * we match another statement
end if
now because * we try again
end if is not a statement, so the <statement> rule *fails*
<?before> sees that the current position is before "end if", so it does nothing
the only thing <?before> does is, sometimes, make matches fail 02:50
it has no effect on anything before or after it
MikeFair But it has left the cursor in the right position for matching on the next part for me 02:51
but that might have been a side effect of other things
sorear <?before> never moves the cursor.
the cursor moved as a result of <statement> _only_ 02:52
MikeFair sorear: It might not be needed anymore, I put it in there because I was having a problem with <statement>* consuming too much 02:53
sorear MikeFair: here's how that works:
MikeFair but that was with a completely different attempt at using a <statement_list> instead of a <statement>*
sorear if you put in <?before> and it *fails*, you backtrack
it forgets the last <statement> and pretends it didn't match after all
and tries again at the previous place 02:54
MikeFair sorear: which in my case would trigger the alternate rule of 'then' <statement> 02:55
but I think I can see why it's not working
dalek rrot/native_pbc2: 8f65204 | rurban++ | / (3 files):
Better bitfiddling with the intel 63 normalize bit

cvt_num16_num10 clears now bit 63 cvt_num10_num16 tries to shift-copy the mantissa ignoring this bit
Test precision regex seems to be stable now.		 02:57
Heuristic branch merge: pushed 196 commits to parrot by rurban
MikeFair sorear: Doesn't seem to work 03:09
rule if_expr {
'if' ~ <?before 'then'> <EXPR>
'then' [ <statement> [ <statement>* 'end' 'if' ]? || 'end' 'if' ]
oh wait, I still have that first before in there
MikeFair tries again 03:10
yeah same problem
'if' <EXPR> 'then' [ <statement> [ <statement>* 'end' 'if' ]? || 'end' 'if' ]
if 1 then say 1; say 2; say 3; doesn't parse 03:11
if 1 then say 1; say 2; say 3; end if does
benabik you want that first to be (if 1 then say 1;) say 2; say 3; ?
MikeFair nods
benabik Why the || 'end' 'if' ? 03:12
MikeFair because sorear said so. :) I think it's explicitly handling the if 1 then end if case
if I read it right
(where there's no <statement> match) 03:13
benabik ah
MikeFair If you're really curious, the original BNF I'm copying is this: 03:15
<ifBlock> =
if <logical> [ <return> ] then {<singleThen> | <return> <multiThen>}
They use 'then' <return> to start a multiThen 03:16
and 'then' <statement> to mean singleThen
where <return> = \\n
benabik What about: 'if' <EXPR> 'then' [ <statement>* 'end' 'if' || <statement> ]
Not terribly efficient, but it should try to match all the statements, then look for an end if, fail to find it then just eat a single statment. 03:17
MikeFair benabik: That's more the approach I was taking, but I don't think that's left factored
but I'll give it a shot
benabik I think trying to do it without a disambiguator is going to be problematic. The original says no return = single, return = multi. 03:18
MikeFair 'if' <EXPR> 'then' [ <statement>* 'end' 'if' || <statement> ] 03:19
same problem
benabik There's no similar disambig in your grammar.
MikeFair if 1 then say 1; say 2; say 3; doesn't parse
MikeFair tries using newlines instead
oh that's what I did hehe
./installable_safire --target=parse -e 'if 1 then say 1 \\n say 2\\n say 3' 03:20
benabik Is the if a rule or token? tokens refuse to backtrack.
MikeFair rule if_expr
benabik \\n inside single quotes doesn't do what you think, I think. 03:21
MikeFair benabik: Well usually if 1 then say 1; say 2; say 3; end if will parse
let's see
yep
./installable_safire --target=parse -e 'if 1 then say 1 \\n say 2\\n say 3\\nend if' parses 03:22
Syntax error at line 3, near "say 2\\nsay "
benabik MikeFair: '\\n' sends \\n to the program, not a newline. You need to use double quotes. 03:23
MikeFair benabik: Oh I'm hitting the <enter> key
not \\n
benabik ah
MikeFair I just didn't want to take up multiple lines in the channel here
Now this might be a problem with statement 03:24
the terminators ; and \\n are actually on "statement_list"
<statement>* %[ ';' | '\\n']
err not '\\n' just \\n 03:25
ok, that seems to be acting better but I can't seem to make the ; || $$ work 03:34
I can get either one of them to work but not a mixture ofthe two... 03:35
I get the same results with either \\n or $$
<ws> is not including \\n 03:36
WooHoo!! I have no idea if it's actually matching the right things, and anything complex takes forever to parse, but at least everything I expect to parse parses! 04:21
rule if_expr { 04:22
'if' <EXPR> 'then'
[ <statement_list> [ 'end' 'if' || 'else' [ <statement_list> 'end' 'if' || <statement> ] ]
|| <statement> ]
}
MikeFair heads off to tuck kids into bed.
:)
04:30 bluescreen_ joined 04:36 eternaleye joined 04:42 particle joined 04:44 eternaleye joined 04:51 benabik joined, benabik_ joined
MikeFair benabik: wb 04:56
05:06 mvorl left 05:35 fperrad joined 07:02 brrt joined 08:09 lucian joined 08:19 Psyche^ joined 10:18 eternaleye joined 10:27 eternaleye joined 11:21 eternaleye_ joined 11:38 eternaleye_ joined 11:43 eternaleye_ joined 12:30 benabik joined 12:41 JimmyZ joined 12:42 mtk joined 12:51 PacoAir joined 13:25 bluescreen joined 14:02 rurban_mobile joined 14:13 benabik joined 14:18 nnunley joined 14:27 rurban_mobile joined
dalek rrot/native_pbc2: b99bb67 | rurban++ | / (2 files):
native numbers: document intel bit63
 14:28
brrt hi #parrot 14:59
15:00 dmalcolm joined
moritz \\o brrt 15:06
brrt \\o moritz
did you perchance ever find out how and if you could load perl6 as a language and a compiler?
moritz brrt: no. I only found out why it doesn't work right now 15:07
(it only calls compreg in the compiling process, but it doesn't emit code for the compreg)
I have to bug jnthn about how to fix that
brrt ok, great, thats good to know :-) 15:08
benabik Hm. Trying to build with Parrot/NQP/Rakudo master give me an error about not finding the 'os' library. I thought we installed a compatibility shim for that? 15:11
Sorry, "trying to build Rakudo HEAD with Parrot/NQP HEAD", is more exactly what I meant.
Coke I know we /talked/ about putting in a shim for that. I don't think I saw a commit (but could have missed it) 15:44
15:46 sivoais joined 16:47 patspam joined 17:25 benabik joined 17:32 benabik joined 17:37 benabik joined 17:40 benabik joined 17:42 benabik joined 17:47 benabik joined
dalek rrot/native_pbc2: 4afbef5 | rurban++ | src/packfile/pf_items.c:
pf_items: keep const qualifiers
 17:47
18:00 benabik joined 18:38 dmalcolm joined
rurban_mobile benabik: we already have an empty dynpmc os for the rakudo problem 18:47
benabik rurban_mobile: Except that I tried to build Rakudo and it failed because it couldn't find library 'os'
18:52 lucian joined
Coke benabik: how did you build rakudo? pre-installed parrot or with -gen-parrot? 18:57
benabik Pre-installed
Built Parrot HEAD, NQP HEAD, Rakudo HEAD
Coke pre-existing older version of parrot installed?
I do see rurban++'s commit for the backward compat. you at d76bef8 or later? 18:59
benabik yeah 19:01
rurban_mobile Yes, d76bef83f04f62f3e5871a6695526eb600b0d7a2 it was. I forgot to include into the original bdw/move-os branch
benabik is at a54a581
rurban_mobile I'm just rescuing my harddisc from the texan weekend power spike which blew up parts of the disc. 19:02
benabik Exact error (while building the setting): error:imcc:loadlib directive could not find library `os' in file '(file unknown)' line 1 19:13
class & 19:19
rurban_mobile benabik: is there a dynpmc/os? 19:23
I could imagine that dynpmc/os did not get installed. 19:25
19:34 contingencyplan joined
Coke benabik: look for the commit that removed it, then the commit that added the stub back, and find the difference. 19:57
rurban++ is probably right about missing something on the install.
20:05 lucian joined
rurban_mobile I'll need about a day to recover my system 20:13
20:54 benabik joined 20:58 benabik_ joined 21:10 lucian joined 22:14 jlaire joined 22:43 whiteknight joined 23:18 GeJ_ joined, benabik joined, perlite joined 23:21 patspam joined 23:27 ingy joined 23:33 mj41 joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!