Closed GoogleCodeExporter closed 9 years ago
For posterity you can see the Chromium side of things here:
https://codereview.chromium.org/169853003/
Original comment by andrewha...@google.com
on 25 Feb 2014 at 6:46
I should also probably note that there were a few changes needed to move from
the old Chromium CLD2 to the latest, so there are some discrepancies between
the Chromium code review and this patch. They are trivial, however, and consist
of:
compile_dynamic.sh:
Uses the latest 0122 tables, and drops the suffix 0715/0122/etc from all
generated binaries it creates
cld2_dynamic_*.cc:
Use 7 tables instead of 6, accommodating the new quad_obj2. Output the
theoretical names of the tables when outputting headers via --head
compact_lang_det_impl.cc:
Same as above, use 7 tables instead of 6 in dynamic mode
Original comment by andrewha...@google.com
on 25 Feb 2014 at 6:49
I just noticed there's a new use of kScoringtables in compact_lang_det_impl.cc:
https://code.google.com/p/cld2/source/browse/trunk/internal/compact_lang_det_imp
l.cc#1976
// Return version text string
// String is "code_version - data_build_date"
const char* DetectLanguageVersion() {
if (kScoringtables.quadgram_obj == NULL) {return "";}
sprintf(temp_detectlanguageversion,
"V2.0 - %u", kScoringtables.quadgram_obj->kCLDTableBuildDate);
return temp_detectlanguageversion;
}
The way this code is written, it SHOULD be compatible with the dynamic chages;
the kScoringtables.quadgram_obj reference should indeed be initialized to null
in dynamic data mode, and that will hit the first if-block and simply return
"". If this is used anywhere critical, that location may need to be patched to
check for isDataLoaded() first. I don't expect this to be a problem, however,
as any code switching to using CLD2 in dynamic mode should be paying close
attention to the data loading work anyways.
Original comment by andrewha...@google.com
on 26 Feb 2014 at 9:51
Patch committed as r151: https://code.google.com/p/cld2/source/detail?r=151
Original comment by andrewha...@google.com
on 3 Mar 2014 at 3:22
Original issue reported on code.google.com by
andrewha...@google.com
on 25 Feb 2014 at 6:30Attachments: