Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-09-01

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:03 sivoais_ joined #marpa
00:04 hobbs joined #marpa
00:04 hobbs joined #marpa
00:33 Aria joined #marpa
00:57 ronsavage1 joined #marpa
01:36 jeffreykegler joined #marpa
02:33 ronsavage1 jeffreykegler: OK if I release MarpaX::Demo::JSONParser now, with your new grammar in it?
04:16 sivoais joined #marpa
04:19 jeffreykegler joined #marpa
04:20 jeffreykegler ronsavage: I take it that it now passes the full suite.  If so, yes, no use hiding this candle under a bushel. :-)
04:20 jeffreykegler Even if Jean-Damien's did turn out to burn a bit brighter. :-)
05:29 ronsavage1 I've released MarpaX::  Demo:: JSONParser V 1.06.
09:56 ronsavage joined #marpa
10:04 ronsavage1 joined #marpa
10:24 ronsavage joined #marpa
11:53 daxim joined #marpa
18:48 jeffreykegler joined #marpa
18:49 shadowpaste "jeffreykegler" at 162.232.214.245 pasted "Timing Libmarpa vs. JSON::XS vs. JSON::PP" (58 lines) at http://scsys.co.uk:8002/420773
18:49 jeffreykegler I've just pasted some more Libmarpa vs. JSON::XS timings
18:54 jeffreykegler Libmarpa is 20% to 40% slower than JSON::XS, depending on how you count things.  There's enough variation in the JSON::XS numbers that you can find pairs of timings where Libmarpa beats JSON::XS.
18:56 jeffreykegler The tests show that the difference in the parser speeds, while measureable, is easily obscured by small differences post-processing and packaging.
18:57 jeffreykegler I also threw in JSON::PP, which Libmarpa always beats by at least 2 to 1.
19:06 jeffreykegler My test code is in the Libmarpa repository: https://github.com/jeffreykegler/libmarpa
19:08 jeffreykegler The purpose is to determine whether Libmarpa is fast enough for production compilers, and IMHO it passes.
19:09 jeffreykegler JSON is hardly a parsed language at all -- it has basically only two structural rules, one for arrays, the other for objects (hashes) -- all the rest is lexing or post-processing.
19:10 jeffreykegler And JSON::XS is a recursive descent parser, carefully hand-coded in C for this very minimal language -- basically I'm entering Libmarpa into the game knowing it's rigged.
19:11 jeffreykegler In this context, the numbers Libmarpa puts up are very acceptable.  In any realistic contest, they'd be much closer.
19:14 jeffreykegler For a large recursive descent parser, if the coding fall short of perfect with respect to efficiency, Marpa could easily put up better numbers.
19:16 jeffreykegler And the result versus JSON::PP shows that Marpa will beat you if you code your recursive descent parser is in anything but C.
22:45 ronsavage1 joined #marpa
23:51 jeffreykegler joined #marpa

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary