Issue 66798

@ sba: Please give it a try.

It seems, targeted to OpenOffice.org 2.2. (See Issue 39992). 

(Unfortunatelly, old language detection was no memory efficient, because it used
all dictionaries during language detection.)

I have tried the text "Rand ist deutsch, margin ist englisch und marge ist
französisch." in OOo 2.3.1 FR with paragraph in german. "margin" and "marge" are
marked by spellchecker and context menu propose English for margin and French
for marge.
If I try the same text in French (Rand est allemand, margin est anglais et marge
est français), Rand and margin are marked by spellchecker and context menu
propose English for both Rand and margin, which is not false but not the best
possible proposition.
I am not sure we can qualify that as defect. So I propose to close this issue as
worksforme.

In the new language detection in m241 it is still not correct. With document
setting and local setting 'German' I do not get French for 'marge'. It seems,
that OOo only looks for the language English and for those languages which are
set as local or document language. But it should look for all installed languages.

tl->regina: 
1) The behavior is according to the spec.
2) You now have the advantage that language guessing can provide you with the
   correct language even if you have no dictionary installed for it.
   (Provided the text is long enough for language guessing to work properly.
   Usually about a sentence is enough.)
3) If one reads the spec it will be seen that the documents languages get
   included as well in the proposal list.
   Thus if you only use some few languages in your document French will become
   available readily if you have at least set one word to French already!
   Also if your keyboard language were French it will be listed from the start.
   I think this it is a very acceptable solution if you only need to actually 
   use a language at least once in order to get that listed as well after...

By pointing out that the behavior is according to spec I do not mean it
can/should not be discussed. I was just meaning if it is being asked to be
changed the issue needs to be labeled as 'enhancement' rather than 'defect.

Is it possible to have a link to the spec?

Here are 2 examples of weird behavior with language guessing:
- http://user.services.openoffice.org/en/forum/viewtopic.php?f=7&t=1995
- http://user.services.openoffice.org/en/forum/viewtopic.php?f=9&t=1369

Is it possible to tweak the configuration file so that guessing language is
limited to only those we want to be guessed?

Basic question implied here is: do we really need OOo to guess the language if
the related dictionary is not installed? If a user opens a document, he should
be aware of the language used inside. Hence, he should be able to deduce that he
needs to install the corresponding the dictionary if not done yet.

cd: Specification of the language guessing feature
http://specs.openoffice.org/appwide/linguistic/Automatic_Language_Guessing_Component.odt

Language guessing is also integrated into this feature:
http://specs.openoffice.org/appwide/linguistic/Set_Language_Attribute_for_Text.odt

*** Issue 66798 has been confirmed by votes. ***

This is a very irritating "feature". I have read the attached documents, but
obviously it is not working as specified. My spreadsheet is defaulting to
English, but I sometimes have cells with Finnish text, usually just a couple of
words long.

I have only the standard OOo3.0 dictionaries installed plus I have added the
Finnish Voikko dictionary. But I continually get told that a misspelled Finnish
text is "Word is Portuguese", or like now "Word is Swahili".

Obviously the language guessing is broken, because it isn't guessing correctly.
And how is it possible to determine "Word is Swahili" if the feature needs a
sentence? - "(Provided the text is long enough for language guessing to work
properly. Usually about a sentence is enough.)"

Many multilingual documents will just have sprinkling of a few foreign words,
not huge blocks of text. In these cases why on earth would I want it to guess
languages (wrongly) which I don't have installed? At the least there should be a
way to disable this or limit it to the installed languages.

tl->jodyfanning: Nothing is broken.
The problem however is that "Set_Language_Attribute_for_Text.odt" is currently
only implemented for the Writer. Calc, Draw, Impress still do it in a different
way (as probably described in the other spec, which I currently can not look
into right now).

So we should either close this issue or adjust the summary to reflect the real
situation. We also should change the issue type to "enhancement" then.

Sorry to come in with as a new and ignorant end user. I am trying to use Calc in
bilingual mode (English and French) and cannot find a way to use the French
spell check for some cells and the English spell check for others (working under
Ubuntu). I have both dictionaries installed and I can use them within the same
document with the Writer. This seems to have to do with the difference in
language handling between Writer and the other components as pointed out by
jodyfanning on 3 February. Others speak about problems with language
guessing/detection (in the title of this issue), but being able to use one's own
selected languages seems more basic. Could the development team advise on
whether there should be two issues or both aspects can be combined here? At any
rate, call it an "enhancement" or a bug, using different languages within the
same document should be given a very high priority, best regards, John

You can use several languages in spreadsheets, just select the cell, and apply a
language to its text in "Format-Cell-Font".

It's just not as comfortable as in Writer where you have the language status bar
control. Thus I name this an enhancement.

Thanks, mba, you have made my day! I am happy now to call the issue of making
language selection and language detection an enhancement. Best regards, John

Glad to read that I could help. :-)

It's absolutely understandable that you didn't find that by yourself.
You are a very good example for my "theory" that treating the language attribute
as a part of the text formatting (and especially as a part of the "Font" tab
page) is a bad design. It's not intuitive and usually people are just used to it. 

IMHO the other things discussed in this issue can be ignored - they either were
misunderstandings or a matter of personal taste. We think that the way that
language guessing and settings works in Writer is as it should be. Please
correct me if I'm wrong.

So I took the liberty to adjust issue type, target milestone and summary
accordingly.

@sba: if you are OK with that, please pass this issue to tl for further
considerations. I don't want to set high expectations, the implementation of
this enhancement is not done in a few days. But maybe we can find a slot of it
in the near future or an interested developer elsewhere. This issue is on our
public "to do" list in the OOo wiki that is also available for the OOo
internship program.

I really think that OpenOffice guessing I am typing Swahili when I only have
English and Finnish dictionaries installed is not a "matter of personal taste".
It seem to be a broken feature. Especially when the detection is only supposed
to work when there is an entire sentence, but it tries when there is only one word.

I would much rather it told me that it was a unknown word than guess it is an
entirely different language that I don't even have installed.

I guess one key thing is that I am talking about Calc. But after trying in
Writer I am not sure that making all the other apps match Writer is that much of
an improvement. The language handling as a whole is quite broken.

For example I have about 4 languages installed. But I only use two. But in order
to get to Finnish (from English) I have to scroll through all the English
variants and all the French variants, along with every other language OOo knows
about before I can get to Finnish. It would be nice that the selection didn't
show languages that are not installed.

It would also be nice if it was possible to order the languages. The shortcut
menu from the status bar only shows one language (the default) at the moment. If
it was possible to sort the languages it would be nice if that showed the first
few from the sorted list.

And for some reason I can't uninstall the French and Spanish dictionaries. The
uninstall option is disabled.

Thanks mba.

I agree that the language selection feature in Writer is very convenient, a
great improvement of that in Calc [Although it is inconvenient that my two
installed dictionaries are not always available without first selecting "Other"
(for example, sometimes I have the choice between US English and French, whereas
I am normally using UK English and French) - have not figured this out yet.]

In Calc, for now I seem to get only either English or French (the two
dictionaries I installed) as default, and can laboriously change for all or part
according to your clarification, thus basically satisfied. I have not yet
experienced the trouble that Calc "guesses" extraneous languages, but agree with
jodyfanning that this could be very inconvenient.

Since I have your ear (sorry if I am bending the rules), I would like to make
some comments on use of Calc as a fresh user:
* Spell check doesn't work in notes
* Difficulty in positioning cursor in notes (several clicks sometimes needed)
* Can't seem to drag value from one cell to a distant one (intermediate cells
also affected)
* Special paste to erase only the note also erases the value
* Have not found how to remove a hyperlink
* System is in general very low in moving and setting pointer and in displaying
values (1 sometimes 2 seconds delay, nothing like in Excel where you get the
feeling of immediate response).
I guess that some or most of these are already intentified, but would appreciate
your general or specific comments. I will be monitoring carefully and
participating in the future. Best regards, John

Being offered first hints from the langages installed, in an order preselected
by the user would be a definite improvement.
Also, having the option to disable hints from other languages would help a lot.
(I personally get a lot of hints from swahili, africaans, and albanian.  But I
understand french and english.)
One way to do this would be to have a list of languages authorised for hints.

Hints from all installed languages could be useful as well.
(i.e., a word could be missing from the language intended, but if it appears
only as a hint from another language, the user will never know that it is
missing from the intended language.  In which case the user would prefer to add
the missing word.)

It would be nice to be able to select 2 (or more) languages simultaneously, as
well.  (e.g. english and some other language for computer-related documents, or
glossaries with descriptions in one language and keywords in anothers.)

Ideally one could select a primary language, with one or more secondary
languages, where words not in the primary language but in a secondary language
would be underlined in, say, blue instead of red.  That of course would be a
more important enhancement.

Note that I use calc much more than writer.