HFST: Bugs

Reported bugs and their resolutions. HFST bugs should be reported to hfst-bugs@helsinki.fi with following information included:

  • Program name and version (use command -V) to find out
  • Steps to reproduce and necessary source files (you may use command script to create log file of shell session)
  • Expected result
  • The platform you used (e.g. use command uname -a)

HFST-lexc does not support regular expressions

Regular expression support is rather preliminary. Some things might work in versions after 6th of April 2009 (or before December 2008).

HFST-lexc regular expressions do not support # as symbol

Versions prior to XXX 2009 in hfst-lexc allow only A-Za-z-_ and Unicode code points using two or more UTF-8 bytes as legal symbols of regular expressions. Work around is to use % before proper fix has been released. E.g.:

< a* # > # ; ! may not work
< a* %# > # ; ! should work

HFST-lexc does not allow 0 in multicharacter symbols

Symbols of style ^A0 will be parsed as ^ A epsilon instead of proper multicharacter symbol, work around by using ^A%0 until fixed. E.g:

Multichar_Symbols  a0
koira0 # ; ! 
koira%0 # !

N.B. the latter may break in newer versions.

HFST-intersect compose does not support weighted transducers

There is some kind of bug in the program, but it hasn't been isolated yet.

HFST-TwolC doesn't understand rules whose centers consist of two set names

If X and Y are sets, X:Y => a _ ; doesn't work. Work around by using variables instead i.e. Vx:Vy => a _ ; where Vx in X Vy in Y;

HFST-Compose-Intersect doesn't warn for missing lexicon output symbols in the rules

If the lexicon contains a word form "abc^M" and the rule alphabet is {a, b, ^M}, the form will be filtered out. HfstComposeIntersect should warn about this. -- MiikkaSilfverberg - 2009-04-10

HFST-TwolC compiles rules X: => a _ ' ; incorrectly where ' is a diacritic.

-- MiikkaSilfverberg - 2009-04-17

HFST-TwolC handle the construction diacritic: incorreclty.

diacritic and diacritic:0 work, though. -- MiikkaSilfverberg - 2009-04-17

Fixed Bugs

HFST-LexC doesn't compile weighted entries.

Fixed since 6th of April 2009

$ cat test.lexc

cat # "weight: 1" ;

$ ./hfst-lexc-static -w test.lexc | hfst-fst2txt -w
0       1       c       c       0
1       2       a       a       0
2       3       t       t       0
3       0

HFST-TwolC compiles /<= rules incorrectly.

The rule a:b /<= b _ ; becomes synonymous with a:b => b _ ;

Fixed in version internal ALPHA-$Revision: 1.10

HFST-Lookup segfaults with input strings which aren't accepted by the transducer.

Fixed in version internal ALPHA-$Revision: 1.4

Topic revision: r8 - 2009-11-03 - ErikAxelson
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback