Sv: "Near Miss" error handling?

"Joachim Durchholz" <joachim_d@gmx.de> skrev i en meddelelse
> Note that Soundex is tied to the English language. Applying it to words
> in any other language produces utter nonsense.
[...]


I think most programmers, regardless of their native language, use English
identifier names. Not all, but most. Besides, variants for other languages
do exist; see SpellCheck from TeraTech, for instance.


> The other point is, if it's mistyped identifiers, you probably don't
> want Soundex anyway. Soundex is useful if you want to identify names
> that were misheard, which errors in programming are (usually) mistyped,
> with totally different error characteristics.
> Typical typos are omitted, transposed, or inserted characters. Typical
> "hearos" are wrong vowels and partly omitted syllables.


Soundex empirically works pretty well for typos (omitted, transposed, etc),
but I'm sure it would need adjustments to be optimal (within its limits) for
identifiers, which have other characteristics than people's names.


Another point is that the number of symbols you have to choose from when
suggesting an alternative gets very limited as you only need to check for
identifers within the scope at hand; be it a module, class, function, etc
(except for external references.)


  - Stein Roger

Return to the comp.compilers page.
Search the comp.compilers archives again.