[
Home
]
[ Index:
by date
|
by threads
]
[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
| Date: | -- (:) |
| From: | Daniel_Bünzli <daniel.buenzli@e...> |
| Subject: | Re: [Caml-list] ANNOUNCE: Xmlm |
Le 27 févr. 07 à 09:28, Stefano Zacchiroli a écrit : > What about the performance? I don't know, I didn't invest a lot of time in profiling, maybe it can be improved. But if you want an unscientific benchmark here you have : I compare the two programs xmllint (which comes on my system) and xmltrip (compiled without any special option) respectively distributed with libxml and xmlm. Note xmllint is a C program so we are not comparing to libxml's ocaml interface. Besides I have no idea how xmllint is written and what it does internally, maybe it does more than xmlm does, so we may well be comparing the uncomparable. The files are <http://www.ximpleware.com/xmls.zip>, uncompressed this is 144 mb of xml files. On macos 10.4.8, G4 1Ghz, 512mo ram. Parse only, without building the tree. > > time ./xmltrip.opt -p -ename ~/tmp/xmls/*.xml > > real 0m53.567s > user 0m51.562s > sys 0m1.043s > > > time xmllint --noent --nocdata --noout --nonet --stream ~/tmp/ > xmls/*.xml > > real 0m25.264s > user 0m24.314s > sys 0m0.725s Parse only, building an in-memory tree. > > time ./xmltrip.opt -t -p -ename ~/tmp/xmls/*.xml > > real 2m2.099s > user 1m44.821s > sys 0m8.215s > > > time xmllint --noblanks --noent --nocdata --noout --nonet ~/tmp/ > xmls/*.xml > real 1m4.590s > user 0m47.561s > sys 0m3.193s Best, Daniel