unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi, I came across thoes tickets and the corresponding thread just yesterday and as fas as I understood the main problem is that newforms ist talking unicode internally and at the interface to the django-ORM. I attached my solution to this problem for django.newforms.models (diffed against latest SVN), which does an encoding to settings.DEFAULT_CHARSET onto the save() between the newforms and django-ORMs. This patch wouldn't be needed or could be removed if django-ORM and/or db-backends all talk unicode/utf-8. Regards, Dirk -- "Feel free" - 5 GB Mailbox, 50 FreeSMS/Monat ... Jetzt GMX ProMail testen: www.gmx.net/de/go/mailfooter/promail-out --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~--- Index: models.py === --- models.py (Revision 4775) +++ models.py (Arbeitskopie) @@ -8,6 +8,7 @@ from forms import BaseForm, DeclarativeFieldsMetaclass, SortedDictFromList from fields import Field, ChoiceField from widgets import Select, SelectMultiple, MultipleHiddenInput +from django.conf import settings __all__ = ('save_instance', 'form_for_model', 'form_for_instance', 'form_for_fields', 'ModelChoiceField', 'ModelMultipleChoiceField') @@ -38,7 +39,10 @@ for f in opts.fields: if not f.editable or isinstance(f, models.AutoField): continue -setattr(instance, f.name, clean_data[f.name]) +try: + setattr(instance, f.name, clean_data[f.name].encode(settings.DEFAULT_CHARSET)) +except: + setattr(instance, f.name, clean_data[f.name]) if commit: instance.save() for f in opts.many_to_many:
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Adrian Holovaty wrote: > Hi Ivan, > > Could you explain again why you think newforms should output > clean_data as bytestrings rather than Unicode strings? I don't think so :-). Quite the opposite. I think it's good that you made newforms in unicode since it effectively gives a start to unicodification and lets us test it early. But there is one issue that is very annoying: when newform saves data to a model instance it fills it with unicode. This creates some issues: - part of the time an object stores bytes (when loaded from db), part of the time -- unicode (when updated from newforms) - __str__s return unicode instead of str - some backends versions can't handle unicode and try to implicitly str() in ascii it with UnicodeEncodeError's The patch that I'm advocating for here does one little thing: when a newform saves an instance it converts unicode into DEFAULT_CHARSET. In other words it makes newforms do decode/encode on both their boundaries: not only on the side facing the web (POSTs, templates) but also on db side. > Are you suggesting that we would convert newforms > clean_data *back* to being Unicode *after* we convert the rest of the > framework to be Unicode-aware? No, clean_data will remain in unicode (and for good). Encoding happens only when clean_data is actually applied to a model instance. > I apologize in advance if you've already brought this up and explained > it. Just trying to understand your thinking here. Yeah, I understand that this thread is rather long :-) --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Adrian Holovaty: > On 2/15/07, Ivan Sagalaev <[EMAIL PROTECTED]> wrote: >> I tried to show that it leaves out only two things: >> >> - if DEFAULT_CHARSET is different than DB charset it won't work (but >> it's a weird situation, most legacy systems have one legacy encoding for >> both) >> >> - it doesn't help if unicode is actually put into models or in raw SQL >> manually but this bug was never about it anyway and won't break anything >> since it fixes newforms, not backends > > Hi Ivan, > > Could you explain again why you think newforms should output > clean_data as bytestrings rather than Unicode strings? The current situation is this: * newforms puts unicode into objects that used to receive only UTF-8 encoded bytestrings * the models (and other parts) only work with bytestrings (or as long as the unicode contains only characters from ASCII) * it is hard to convert all the rest of Django to be able to deal with unicode and bytestring at the same time, and it seems that this has been postponed until after 1.0. Ivan proposes a fix that tries to convert unicode to bytestrings at the boundary of newforms by encoding unicode to bytestrings in clean_data. (I have not checked whether this resolves all or at least a big part of the problem, and I don't have a position about this, yet.) It looks like a step backwards, but as long is we don't try to make everything unicode compatible at the same time, we need to encode the unicode strings at some boundary or the other. Is clean_data the right point for this? Michael -- noris network AG - Deutschherrnstraße 15-19 - D-90429 Nürnberg - Tel +49-911-9352-0 - Fax +49-911-9352-100 http://www.noris.de - The IT-Outsourcing Company --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On 2/15/07, Ivan Sagalaev <[EMAIL PROTECTED]> wrote: > I tried to show that it leaves out only two things: > > - if DEFAULT_CHARSET is different than DB charset it won't work (but > it's a weird situation, most legacy systems have one legacy encoding for > both) > > - it doesn't help if unicode is actually put into models or in raw SQL > manually but this bug was never about it anyway and won't break anything > since it fixes newforms, not backends Hi Ivan, Could you explain again why you think newforms should output clean_data as bytestrings rather than Unicode strings? If I understand your argument correctly, you're saying newforms should be rolled back to bytestrings because the rest of the framework isn't Unicode-aware yet. Are you suggesting that we would convert newforms clean_data *back* to being Unicode *after* we convert the rest of the framework to be Unicode-aware? I apologize in advance if you've already brought this up and explained it. Just trying to understand your thinking here. Adrian -- Adrian Holovaty holovaty.com | djangoproject.com --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > (meeting postponed ...) Nice :-) > You're right, sorry. I was in a different ticket and somehow thought > it was the same. > > Yes, #3370 looks interesting and is a different solution. I'm not > sure whether it deals with all the issues of this thread. I tried to show that it leaves out only two things: - if DEFAULT_CHARSET is different than DB charset it won't work (but it's a weird situation, most legacy systems have one legacy encoding for both) - it doesn't help if unicode is actually put into models or in raw SQL manually but this bug was never about it anyway and won't break anything since it fixes newforms, not backends --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Ivan Sagalaev: > Michael Radziej wrote: >> Ivan Sagalaev: >> >>> Michael, the ticket http://code.djangoproject.com/ticket/3370 just got a >>> patch that does a) and it's really small. It's not as full as having b) >>> and d) but I think they are really a corner cases: b) for different >>> encodings in DB and in web, d) for handling unicode input to DB backend >>> *without* newforms. >>> >>> In other words I think that patch is just right for current situation >>> because it fixes the bug for people trying to use newforms now. I'm +1 >>> on just committing it as is. >> I'm not sure if the fix is on the right level. StrAndUnicode is used >> in a lot of places. Is it sure that it won't put xmlcharref-encoded >> data into the database? I only had a very quick look on it (and I >> need to go to a meeting now). > > Uhm... Are we talking about the same patch? This is it: > http://code.djangoproject.com/attachment/ticket/3370/models.py.diff > > It doesn't mention StrAndUnicode at all. (meeting postponed ...) You're right, sorry. I was in a different ticket and somehow thought it was the same. Yes, #3370 looks interesting and is a different solution. I'm not sure whether it deals with all the issues of this thread. Michael -- noris network AG - Deutschherrnstraße 15-19 - D-90429 Nürnberg - Tel +49-911-9352-0 - Fax +49-911-9352-100 http://www.noris.de - The IT-Outsourcing Company --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > Ivan Sagalaev: > >> Michael, the ticket http://code.djangoproject.com/ticket/3370 just got a >> patch that does a) and it's really small. It's not as full as having b) >> and d) but I think they are really a corner cases: b) for different >> encodings in DB and in web, d) for handling unicode input to DB backend >> *without* newforms. >> >> In other words I think that patch is just right for current situation >> because it fixes the bug for people trying to use newforms now. I'm +1 >> on just committing it as is. > > I'm not sure if the fix is on the right level. StrAndUnicode is used > in a lot of places. Is it sure that it won't put xmlcharref-encoded > data into the database? I only had a very quick look on it (and I > need to go to a meeting now). Uhm... Are we talking about the same patch? This is it: http://code.djangoproject.com/attachment/ticket/3370/models.py.diff It doesn't mention StrAndUnicode at all. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > A few days ago, I wrote: >> I see three ways to fix the problem in #3370: >> >> a) newforms stops passing unicode strings to the Database API and uses >> bytestrings. >> >> b) the database wrapper in Django sets connection.charset (but needs to >> translate the charset name since the databases don't understand all >> charset name variants, see ticket #952 here). This is the approach of >> the patches in tickets #1356 and #3370. >> >> c) the database wrapper in Djago must check whether it gets unicode. In >> this case, it needs to encode it into a bytestring. > > I now see a fourth way that would resolve #952 at the same time: > > d) make the database wrapper accept both unicode and bytestrings in > the models, but always pass unicode strings to the database backend. Michael, the ticket http://code.djangoproject.com/ticket/3370 just got a patch that does a) and it's really small. It's not as full as having b) and d) but I think they are really a corner cases: b) for different encodings in DB and in web, d) for handling unicode input to DB backend *without* newforms. In other words I think that patch is just right for current situation because it fixes the bug for people trying to use newforms now. I'm +1 on just committing it as is. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On Feb 1, 4:16 pm, Michael Radziej <[EMAIL PROTECTED]> wrote: > Ivan Sagalaev: > > > Michael Radziej wrote: > >> d) make the database wrapper accept both unicode and bytestrings in > >> the models, but always pass unicode strings to the database backend. Sounds like a reasonable proposal. You may even consider logging deprectation messages in the case of bytestrings appearing in models (but be careful not to create a flood of these). --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On Jan 27, 2007, at 6:44 PM, ak wrote: > And another thing I still don't understand is: let's pretend I use > MySQL 4.0 with national charset and my templates are in the same > charset too. How would work: It should not work. -- Julian 'Julik' Tarkhanov please send all personal mail to me at julik.nl --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On Jan 27, 2007, at 6:44 PM, ak wrote: > 1. newforms are with unicode inside > 2. ORM is with str inside 3. welcome to the world of pain -- Julian 'Julik' Tarkhanov please send all personal mail to me at julik.nl --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi, A few days ago, I wrote: > I see three ways to fix the problem in #3370: > > a) newforms stops passing unicode strings to the Database API and uses > bytestrings. > > b) the database wrapper in Django sets connection.charset (but needs to > translate the charset name since the databases don't understand all > charset name variants, see ticket #952 here). This is the approach of > the patches in tickets #1356 and #3370. > > c) the database wrapper in Djago must check whether it gets unicode. In > this case, it needs to encode it into a bytestring. I now see a fourth way that would resolve #952 at the same time: d) make the database wrapper accept both unicode and bytestrings in the models, but always pass unicode strings to the database backend. Details: For #952 to work, the name of the character encoding has to be translated from python naming conventions to these of the used backend, and this would need a huge table (see the ticket). It looks easy, but it's a major annoyance. Now, instead of doing this, how about modifying the database wrapper so that it actually tests whether it gets unicode or bytestrings, and in the case of bytestrings, decodes it to unicode using settings.CHARACTER_SET as encoding? Then it could use unicode to talk to its backend. As far as I see, psycopg2 is unicode capable, and python-MySQLdb, too. This is different from the proposal in the thread 'Unicode or Strings in Models', as I'd still accept both forms in the model and deal with it only when I send it to the database. 'Only unicode in models' would be a major change with many scattered pieces. My proposal is for a transition phase, to support piece-wise conversion to Unicode without breaking everything on the way (as newforms does). Disadvantage: The backend will probably decode it again to get it across the wire, to either UTF-8 or settings.DEFAULT_CHARSET (or something else), adding overhead to the database communication. I think this is a necessary transition from bytestrings to the Great Unicodification of Everything. As soon as there's unicode everywhere, the code that deals with bytestrings can be removed and the solution will fit in perfectly. What do you think? Michael -- noris network AG - Deutschherrnstraße 15-19 - D-90429 Nürnberg - Tel +49-911-9352-0 - Fax +49-911-9352-100 http://www.noris.de - The IT-Outsourcing Company --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Bill de hOra wrote: > Yep; it's a problem on the way back in. Python won't let you interpolate > encoded bytestrings and unicode; you have to state the encoding. Ivan - > could the db encoding be declared in settings.py? This is what #952 is about. Though it doesn't convert things for DB on Django side, it declares Django's data encoding to DB instead so it can convert. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Ivan Sagalaev wrote: > Michael Radziej wrote: >> >> I don't see a problem with the generic views since they pass bytestrings >> to the database wrapper, this gets as bytestrings to MySQLdb, and for >> bytestrings the charset attribute is not used at all. > > Umm... This is the exact problem with byte strings: that they require > knowledge of a charset somewhere. Yep; it's a problem on the way back in. Python won't let you interpolate encoded bytestrings and unicode; you have to state the encoding. Ivan - could the db encoding be declared in settings.py? cheers Bill --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > I thank you for all your patience with me. I was completely off-track. I > read all the mails again, and everything is starting to make sense now. Then I hope not to confuse you (and everyone else) with my answer :-) > First, contrary to my former opinion, #3370 is a bug in the newforms > module, as it is passing unicode to the database API which is not ripe > for it and will break as soon as you leave ASCII. I wouldn't call it a bug. Newforms are intended to work in unicode. They don't play nice with db backends now but it's a question what should be changed: newforms to supply byte strings or db backends to accept unicode. > I see three ways to fix the problem in #3370: > > a) newforms stops passing unicode strings to the Database API and uses > bytestrings. > > b) the database wrapper in Django sets connection.charset (but needs to > translate the charset name since the databases don't understand all > charset name variants, see ticket #952 here). This is the approach of > the patches in tickets #1356 and #3370. > > c) the database wrapper in Djago must check whether it gets unicode. In > this case, it needs to encode it into a bytestring. I believe option a) and b) together will do the work. Now we have all these confusing bugs because db backends receive two kind of inputs: unicode from newforms and byte strings from oldforms (a majority of existing code I think). Newforms are now "guilty" of introducing unicode into party so I think it's better to keep all the conversions there. Option b) is needed because a db backend should know in which single-byte encoding it receives data. The great advantage of unicode is that you shouldn't supply a text's language alongside, it's encoded right there. But with byte strings it's necessary. Option c) scares me :-). Because the need in working with byte strings (and hence in options a) and b)) remains but also introduces an ability to accept but not to issue unicode objects also. I don't think people would thank us for this :-) > With all three variants, what encoding should be used? We currently > issue (without #952) a 'set name utf8' at the beginning of each > connection, so the database server expects to receive utf8. So, > shouldn't we currently always use utf8 encoding, regardless of what is > in settings.DEFAULT_CHARSET? No we shouldn't. In fact this was never working properly, #952 is an old bug. It kinda works most of the time because the default value of DEFAULT_CHARSET is 'utf-8' and most apps don't change it. But if they do and actually work with non utf-8 data then when fed into database declared as utf-8 they will break because an arbitrary single-byte encoding is not well-formed utf-8. Databases react differently: Postgres complains that it's not utf-8 and refuses to accept garbage (I love Postgres :-) ). MySQL, at least some versions, just won't check the encoding and store data as a byte array. Sorting and case insensitivity won't work but at least you can SELECT everything back unchanged which supports the notion that it "works" :-). Actually this means that #3370 is safe to include because it's MySQL-only, doesn't affect byte strings at all because of MySQL's liberal interface and actually fixes a bug when it receives unicode from newforms. I'm against it only because it creates this incomprehensible mess of conventions and edge cases neutralizing each other... #952 is just a more general way of doing things. > Well, the current patch in #3370 (I still ignore __repr__) only changes > the charset attribute of a connection, and this attribute is used only > to encode unicode strings when sending data to the database, or to > decode bytestrings received from the database when MySQLdb is configured > to produce unicode ('use_unicode'). BTW I'm -1 on switching backends to unicode right now because: 1. We should manually decode/encode for backends that can't do it (say, psycopg1) 2. We immediately get __str__'s returning unicode objects which will open a can of worms of confusions (and flame wars :-) ). > I don't see a problem with the generic views since they pass bytestrings > to the database wrapper, this gets as bytestrings to MySQLdb, and for > bytestrings the charset attribute is not used at all. Umm... This is the exact problem with byte strings: that they require knowledge of a charset somewhere. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi there, I thank you for all your patience with me. I was completely off-track. I read all the mails again, and everything is starting to make sense now. This is going to be a lengthy email about #1356 and #3370, but please do read until the end. Short executive summary: It's really a bug, and the patch is not bad, but incomplete. First, contrary to my former opinion, #3370 is a bug in the newforms module, as it is passing unicode to the database API which is not ripe for it and will break as soon as you leave ASCII. #3370 is independent of #952. I see three ways to fix the problem in #3370: a) newforms stops passing unicode strings to the Database API and uses bytestrings. b) the database wrapper in Django sets connection.charset (but needs to translate the charset name since the databases don't understand all charset name variants, see ticket #952 here). This is the approach of the patches in tickets #1356 and #3370. c) the database wrapper in Djago must check whether it gets unicode. In this case, it needs to encode it into a bytestring. With all three variants, what encoding should be used? We currently issue (without #952) a 'set name utf8' at the beginning of each connection, so the database server expects to receive utf8. So, shouldn't we currently always use utf8 encoding, regardless of what is in settings.DEFAULT_CHARSET? This point has caused a lot of confusion. Ivan wrote: > I'm -1 on setting MySQL connection to 'utf8' in #3370. It *will* make > sense when we will have newforms ready and models containing unicode. > But now most of Django is a byte string country. A bright example are > generic views that take data from web and store it to models without any > conversions. This patch will feed 'windows-1251' or 'iso-8859-1' to > MySQL saying that "it's utf-8" and MySQL will try to convert it and most > certainly will store just strings of ''. Well, the current patch in #3370 (I still ignore __repr__) only changes the charset attribute of a connection, and this attribute is used only to encode unicode strings when sending data to the database, or to decode bytestrings received from the database when MySQLdb is configured to produce unicode ('use_unicode'). Here's what the documentation in MySQLdb-1.2.2b2 says: use_unicode If True, CHAR and VARCHAR and TEXT columns are returned as Unicode strings, using the configured character set. It is best to set the default encoding in the server configuration, or client configuration (read with ==> read_default_file). If you change the character set after ==> connecting (MySQL-4.1 and later), you'll need to put the ==> correct character set name in connection.charset. If False, text-like columns are returned as normal strings, but you can always write Unicode strings. *This must be a keyword parameter.* (But, the charset parameter is also used when you pass in unicode without setting use_unicode) python-MySQLdb-1.2.1p2 is similar, only that there it is no keyword parameter. There's an interesting difference between 1.2.1p2 and 1.2.2b2: For 1.2.1p2, you have to change the charset attribute of the existing connection. If you try this on 1.2.2b2, it won't work. For 1.2.2b2, you either have to pass a 'charset' parameter when you create the connection, or you can call a method set_character_set(). Both of these won't work for 1.2.1p2, of course :-( So, the APIs of python-MySQLdb are incompatible with each other (within a minor version change!) This explains the differences between #1356 and #3370. We need a patch that plays well with both versions of python-MySQLdb. I don't see a problem with the generic views since they pass bytestrings to the database wrapper, this gets as bytestrings to MySQLdb, and for bytestrings the charset attribute is not used at all. Of course, as soon as #952 has been applied, we need to use the encoding from settings.DEFAULT_ENCODING. Michael P.S.: If you set the charset parameter in 1.2.2b2's Connection.__init__(), the default for use_unicode will be True, and python-MySQLdb will return unicode strings. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
[EMAIL PROTECTED] wrote: > I think the next step in the unicodeification of django is to decide where > the conversions happen. Or has this already been decided? > > I like the picture of "unicode circle of trust": everything inside the circle > is trusted as unicode strings. Everything outside has to be encoded/decoded. > It's pretty clear the database is outside, the http gets/posts are outside > too. But what about templates? What about settings/views/models? > I guess if that is decided, we can have a "unicode roadmap". I guess there > are a few people who have spare time and knowledge to help django become > unicode. Yes this was discussed and resolved pretty much like you described: everything is in unicode except the Web and the database. The roadmap is here: http://code.djangoproject.com/wiki/UnicodeInDjango --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > Hey, I now finally understand why you need #952 as soon as you switch to > a different charset. I understand your point, but I'd rather offer a > solution than postponing this for such a long time. +1 #952 is good to include now since it plays nice with byte string models that we have now. Newforms issue, that they can't be automatically dumped to models, is also really a separate thing. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi, Ivan Sagalaev schrieb: > Michael Radziej wrote: >> I'm not sure about what the last sentence means--are you suggesting to >> put #3370 (the mysql part) into "Needs design decision"? > > ## 3370 > > I'm -1 on setting MySQL connection to 'utf8' in #3370. It *will* make > sense when we will have newforms ready and models containing unicode. > But now most of Django is a byte string country. A bright example are > generic views that take data from web and store it to models without any > conversions. This patch will feed 'windows-1251' or 'iso-8859-1' to > MySQL saying that "it's utf-8" and MySQL will try to convert it and most > certainly will store just strings of ''. The patch is working for > the author only because it feeds newforms' unicode objects right into > models which is wrong (we hadn't convert models to unicode yet). Ah, I see. Somehow it wasn't clear to me that POSTs and GETs are just passed along, but now that you mention it, it looks so obvious. Thanks, that was the missing piece that kept me from proper understanding. > But the __repr__ part is plain incorrect: Now, let's keep __repr__() apart, it's a different issue. We can come back to it later. > ## 952 > > This patch tries to set connection encoding to the one used for web: > DEFAULT_CHARSET. But when we convert Django to unicode (we'll have to do > it anyway because of newforms) this won't be necessary because models > will be unicodified too. Then it'll make sense to set 'utf8' in all > backends as a connection encoding. Hey, I now finally understand why you need #952 as soon as you switch to a different charset. I understand your point, but I'd rather offer a solution than postponing this for such a long time. > ## Suggestion > > Now I think we should close all these bugs. Don't laugh (or cry)! #952 > is neither long-term nor helps ak's case, #3370 is broken (sorry, ak, > but it is) and #1356 is a dupe of #3370. I agree to close #1356 and #3370, but #952 seems to be valuable independent of ak's case. I'd rather put #952 into "Needs design decision", because that's really the realm of the core to decide, but it looks a bit that Adrian has already accepted it (as he reopened it). SmileyChris, you did the initial triage on #952, do you read me? What's your opinion here? As I said, this is __repr__() kept aside. Let's tackle it after the connection encoding. Michael --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Ok, thanks for that Ivan, Michael - ignore what I said before :-). The real question, then, is what will it take to get Django unicode uh, "safe" (not sure if that's the best term) before 1.0. I realise that this looks like it's going to be fairly major to sort out, but if we don't then we're going to have all sorts of irritating little bugs like these ones popping up repeatedly. --Simon --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > I'm not sure about what the last sentence means--are you suggesting to > put #3370 (the mysql part) into "Needs design decision"? ## 3370 I'm -1 on setting MySQL connection to 'utf8' in #3370. It *will* make sense when we will have newforms ready and models containing unicode. But now most of Django is a byte string country. A bright example are generic views that take data from web and store it to models without any conversions. This patch will feed 'windows-1251' or 'iso-8859-1' to MySQL saying that "it's utf-8" and MySQL will try to convert it and most certainly will store just strings of ''. The patch is working for the author only because it feeds newforms' unicode objects right into models which is wrong (we hadn't convert models to unicode yet). But the __repr__ part is plain incorrect: try: return '<%s: %s>' % (self.__class__.__name__, self) except UnicodeEncodeError: return '<%s: %s>' % (self.__class__.__name__, self.__str__().encode(settings.DEFAULT_CHARSET)) The __str__().encode(...) is wrong because it's already 'str' and you can't encode it any further. It was working for patch author because he had __str__ of a model returning a unicode object. It's wrong and it should be fixed after the whole unicodfication of Django. But patching it this way will break perfectly normal code where people don't assign unicode objects to model properties. Granted, the breakage won't be very bad because people don't show __repr__ to users often. But it's still bad. ## 952 This patch tries to set connection encoding to the one used for web: DEFAULT_CHARSET. But when we convert Django to unicode (we'll have to do it anyway because of newforms) this won't be necessary because models will be unicodified too. Then it'll make sense to set 'utf8' in all backends as a connection encoding. ## Suggestion Now I think we should close all these bugs. Don't laugh (or cry)! #952 is neither long-term nor helps ak's case, #3370 is broken (sorry, ak, but it is) and #1356 is a dupe of #3370. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
ak wrote: > Bjorn, if you read my first messages and specially my patch #3370, you > find that I made a suggestion that if the guys want to move to unicode > they better drop all native encodings support and so does my patch. With all due respect, you seem to not understand this. 'Unicode' does not mean 'dropping native encodings support'. This is just FUD. Your patch in #3370 is broken (as I showed to you in personal mail) because it 'encodes' __str__ which works only for your special case where you assign a unicode object to a model property and return it from __str__. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi Simon, [EMAIL PROTECTED]: > +1 > > I was just coming to the same conclusion - #952 looks good to go, and > #3370 could be split into the __repr__ and mysql issues. __repr__ and > #952 are easy to solve. The rest of it needs the cores to come to a > decision about this. I'm not sure about what the last sentence means--are you suggesting to put #3370 (the mysql part) into "Needs design decision"? Michael --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
+1 I was just coming to the same conclusion - #952 looks good to go, and #3370 could be split into the __repr__ and mysql issues. __repr__ and #952 are easy to solve. The rest of it needs the cores to come to a decision about this. --Simon --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi, ak schrieb: > After some thoughts I came to the following conclusion: if you guys > want to keep support of legacy charsets in fact you don't have to > force model objects too be unicoded. Firstly, they are passed to > templates and filters and we can't mix legacy charsets with unicode in > one template. Next, if I don't use unicode, I don't have to code my > python sources (views) in unicode. So, I need to be able to pass > string values into my model objects and my strings are not unicoded. > > So if everyone agreed, the way is simple: > 1. when django loads data from db and fills in a model object, all > strings have to be encoded according to DEFAULT_CHARSET > 2. when django passes data from form object to model object, it has to > encode strings according to DEFAULT_CHARSET again This thread is moving more and more away the tickets. I started it to get some help in deciding how to proceed with these ... Regarding ak's proposal, this is going against a widely shared agreement within the python world that applications should internally use unicode strings (not: utf8 strings) and decode/encode to a bytestring at the boundaries, which is usually input/output, or for database applications it's the communication between the database backend (e.g. MySQLdb) and the database. I'm not in a position to make any decisions for django, but I'm pretty sure that you cannot convince the core developers to follow your path. Down to earth and back to tickets, my current understanding is this: The problem that started the original thread in django-users was that the MySQLdb backend thought it was using latin-1 encoding for the connection and therefore could not encode '', which is in iso-8859-15 but not in iso-8859-1 aka iso-latin-1. Ticket #2896 seems to explain how this can happen. In my opinion, each of the three tickets in the subject should solve this issue, and none tries to cope with templates written in a different encoding than settings.DEFAULT_ENCODING. #952 allows to use a different encoding on the connection than settings.DEFAULT_CHARSET. It does it for all backends. #1365 sets connection.charset in the mysql backend to utf8. This makes the MySQLdb use utf8 encoding, but it's hackish and has been reported not to work in all environments. #3370 opens the mysql backend connection with charset='utf8', which seems a cleaner way to do the same as #1365. It also fixes the __repr__ of models (not sure if this is the best way, but this can be added to any of the other patches) My bottom line is that #952 has a different scope than the other two tickets, and that #1365 should be closed as duplicate of #3370. #3370 and #952 can co-exist. So, would anybody object against closing #1365 and promoting #952 and #3370 to "Accepted" (which was their state before we started this discussion)? Michael --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On Jan 28, 2:02 pm, "ak" <[EMAIL PROTECTED]> wrote: > Bjorn, if you read my first messages and specially my patch #3370, you > find that I made a suggestion that if the guys want to move to unicode > they better drop all native encodings support and so does my patch. You mean require all I/O edge/boundary points to convert to/from Python unicode strings? (We'll of course need to support non-UTF character encodings in databases, files, the web, etc.) > Then people started to answer me that this is wrong. And at the moment > noone is able to explain the whole thing and answer my quesions: > 1. how do they want to support templates and python code (views/ > scripts) in native encodings if django itself would be all in unicode. > The only way i see is to encode/decode everything at programmer's end > and this means for me no native encodings support at all. Support for Unicode strings (u"") in code is described in PEP-263, e.g., #!/usr/bin/python # -*- coding: -*- Unfortunately it's not implemented yet (AFAIK), so you can't just have unescaped literals: s = u"encoded text goes here" # doesn't work yet; pending PEP-263 An alternative for literals in code is to surround them with unicode() and specify the appropriate encoding: s = unicode("encoded text goes here", "encoding name") An even better way is to externalize all strings in .po files and use gettext, which has some support for returning unicode strings. I guess templates could have their character encoding identified either through a similar mechanism, through a global settings variable, or just use the system default encoding. > 2. how do they want to support legacy databases if db connection speaks > unicode I'm not sure I can follow you. How to configure a database adapter depends on the database and adapter you're using. Some can accept unicode strings; for those that don't I guess you'll need a wrapper of some sort. Rgds, Bjorn --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
After some thoughts I came to the following conclusion: if you guys want to keep support of legacy charsets in fact you don't have to force model objects too be unicoded. Firstly, they are passed to templates and filters and we can't mix legacy charsets with unicode in one template. Next, if I don't use unicode, I don't have to code my python sources (views) in unicode. So, I need to be able to pass string values into my model objects and my strings are not unicoded. So if everyone agreed, the way is simple: 1. when django loads data from db and fills in a model object, all strings have to be encoded according to DEFAULT_CHARSET 2. when django passes data from form object to model object, it has to encode strings according to DEFAULT_CHARSET again In fact, my patch #3370 is wrong then, actually newforms.model.save() method should be patched to recode clean_data from unicode to DEFAULT_CHARSET (if it differs) when passing this data to model object and for now we would get everything in place: utf8-based templates and legacy-charset-based templates would be both correctly supported and any national characters would be stored in db perfectly as they do now with oldforms (ofcourse remember what I said about #952) And the second required patch is about recoding unicode strings loaded from db to DEFAULT_CHARSET (if differs) when passing them to model objects and back from DEFAULT_CHARSET to unicode when we save model objects to db. This patch will solve #952 issue and again it will work ok with both unicode and legacy-charset based templates. And even more here: if we have a legacy database which doesn't understand unicode, we can realize this fact immediately after connecting to db and decide the correct way to decode/encode strings. As I see, this way fixes all unicode/charsets issues and answers all questions. So, if there are no objections, I can write this patch tomorrow or by monday. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > 1. Are all these tickets really about the connection encoding? > > 2. If so, what's the problem of using utf8 for the connection for > everybody? I don't see how this would be a problem for anybody who is > using a different encoding for templates, within the database's storage > or else, since there's no loss in converting anything into utf8. Or is > there? I agree with the 2nd point. You still can run into a theoretical problem with it in a scenario when an input is richer than a storage: - a database that is internally stores data in a legacy encoding (say iso-8859-1) - a web frontend that talks utf-8 - a user enters, say, Russian characters into a form - data travels as utf-8 right until db where it will fail to encode them in iso-8859-1 because it doesn't have place for Russian characters But it's indeed a very theoretical case. Most legacy system use the same legacy encoding for both backend and frontend and there would be no errors in the path: legacy (web) - unicode (newforms) - utf-8 (db connection) - legacy (db) --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
ak wrote: > Could someone please explain me what was a problem with unicode support > in oldforms so newforms have been made with unicode inside ? I can! The thing is it has absolutely nothing to do with forms, it's just historical coincidence. Originally Django was written with using byte strings everywhere and there were no such thing as "conversion problem". However there were problems with incorrect string operations on byte strings (maxlength counting, upper/lower casing, etc.) Some time ago there was a decision to convert Django to work internally with unicode strings and convert them into byte strings on boundaries to the web and to the database. And there were no such thing as newforms at that moment. And then Adrian started to implement newforms and he has chosen to do its internal in unicode, for compatibility with Django's future as I understand it. > Kick me if I wrong but what is a real reason to convert bytes back and > forth ? Religion ? Reasons are purely technical... I'll list them but please do read until the end of the letter before you disagree. I believe you just misunderstand some things about unicode. 1. Unicode is a universal encoding that can store all characters. Without universal encoding an app written by a Russian programmer wouldn't be able to use a library written by a French programmer. This is why we need unicode. 2. In Python unicode strings can be either 'unicode' objects or byte-strings encoded in utf-8. The problem with utf-8 is that you can't string operations with it. For example you can't cut a month's name to 3 letter just by doing month[0:3] because letters can occupy different count of bytes. This is what unicode objects are for and why Django internally should work with unicode. May I recommend you my post about unicode and bytes (it's in russian): http://softwaremaniacs.org/blog/2006/07/28/unicode-and-bytes/ > I agree with everyone who says that unicode is a > must and 'legacy' charsets are crap but guys I already have a BIG > application that was about 80% migrated from other python frameworks to > django some time ago and for legacy reasons it was all in national > charset, not unicode. What gives you an idea that Django won't work with this data? All this unicode stuff is purely internal. If you want your app to output windows-1251, set DEFAULT_CHARSET to windows-1251 and data would be automatically converted from and to it. I believe even newforms already use this setting to convert unicode data for templates (if not it should be just fixed and I'm happy to make a patch since I got some free time). > Then I found that oldforms support will be > dropped soon or later. So we at here have decided to start moving (yes, > moving again !!!) all our code to newforms and what we got ? We got > that we now have to recode everything to utf-8 Sure not :-). I'd say it would be wise thing to do *eventually*. But for now you absolutely can keep your templates and python sources in windows-1251. > Did anyone who used unicode with oldform has any problems ? I am sure > noone did. In fact nobody used unicode with old forms. All things in request.POST, manipulator.flatten_data and in db models were always in byte strings (except db models with psycopg2). And there were problems with it. They were just fixed very early (a couple of them by yours truly). > So guys please explain me what was a reason to make me to migrate to > unicode ? I still think that you're confusing migrating Django internals to unicode objects and converting your files to utf-8. It's not about the latter. > My opinion is simple: let's decide once ether django is for unicode or > django supports both unicode and national charsets and then let's work. Sure Django does and will support national charsets. This is why we have DEFAULT_CHARSET setting. Internal unicode just lets Django have all the encode/decode stuff localized in two places instead of littered all over the code. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael, of you read again the topic about euro sign in newforms you can find that this touches everything. Personally I couldn't find a way to use utf-8 to connect MySQL and keep using cp1251 in my templates: it basically doesn't work. With my patch (#3370) and utf8 everywhere it does. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Guys Could someone please explain me what was a problem with unicode support in oldforms so newforms have been made with unicode inside ? Kick me if I wrong but what is a real reason to convert bytes back and forth ? Religion ? I agree with everyone who says that unicode is a must and 'legacy' charsets are crap but guys I already have a BIG application that was about 80% migrated from other python frameworks to django some time ago and for legacy reasons it was all in national charset, not unicode. Then I found that oldforms support will be dropped soon or later. So we at here have decided to start moving (yes, moving again !!!) all our code to newforms and what we got ? We got that we now have to recode everything to utf-8 and search for bugs in over than 10k lines of our oldforms-based code until we move everything to newforms and utf-8. But really why ? Did anyone who used unicode with oldform has any problems ? I am sure noone did. Did anyone who used native encodings with oldforms has any problems (except of patch against one line of code I dscribed before or #952) ? Noone did. So guys please explain me what was a reason to make me to migrate to unicode ? Django is a web framework for perfectionists with deadlines. I see may perfectionists here but what about deadlines ? My opinion is simple: let's decide once ether django is for unicode or django supports both unicode and national charsets and then let's work. If you tell me that from now there is only "unicode future" i'd agree and start searching for bugs and sending patches like #3370 --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On Jan 26, 2007, at 11:47 AM, Michael Radziej wrote: > # 1356 sets the charset attribute of the mysql backend connection to > 'utf8' for mysql version >= 4.1 And leaves everyone who wants to operate in 8 bits out in the cold. Where they actually ought to be anyway, but I tried to stay liberal in 952 - primarily because it's still unknown how Django authors want to approach this. -- Julian 'Julik' Tarkhanov please send all personal mail to me at julik.nl --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
On Jan 26, 2007, at 2:25 PM, Gábor Farkas wrote: > > Julian 'Julik' Tarkhanov wrote: >> >> >> Python's unicode is actually UTF-16 > > on linux it's usually utf-32, and on windows it's usually (always?) > utf-16. sorry I forgot that - it's been a year at least since I last touched Python (actually it was for the Django test drive) > > but you should not care about it. you see, in python, > the unicode-strings are a separate data-type, and there's > just no way to take a bytestring, and tell python: "from now on, > you are an unicode-string, because i know that you are encoded in > utf-16." segregating ustrings and strings is BBD, been' telling it for years. The latest I heard is that the next major Py will abolish bytestrings for good. Getting back to the issue that we were on, I am still strongly advocating the "don't go there" approach for anything but Unicode. How it should be handled in relation to source code is unknown to me (AFAIK Python has a pre-amble sort of declaration that you can actually use to tell the interpreter which encoding your source is in). I just know you hit some major pain when you expect ustrings and get bytestrings instead (and in Python, just as in Perl, only about 30% of the libraries actually care about what they give you). > so while it might be, that the conversion from utf-16-bytestrings to > unicode is sometimes faster thatn converting from utf-8-bytestrings to > unicode, you can't be sure, because as i wrote above, the internal > unicode-encoding is not fixed. > >> whereas IO and the databases mostly >> speak UTF-8 - >> so no, you can't dump it over the wire. > >> We Rubyists are a tad happier >> because we now >> have all in UTF-8 > > you mean that regexes, and all the methods of the string-class now are > unicode-aware in ruby? :) Regexes are unicode-aware for some time already except the case- sensitivity and the class repertoire (which will be fixed when Oniguruma is there). As for the string methods, we mostly took care of them with AS::Multibyte (without silly subclassing) and that works wonders for me. The greatest advantage is that I never have to check what's coming down the pipe because there's only one String to rule them all. -- Julian 'Julik' Tarkhanov please send all personal mail to me at julik.nl --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Julian 'Julik' Tarkhanov wrote: > > > Python's unicode is actually UTF-16 sorry, but no. it's not utf-16. it's decided at compile-time, and i'ts either utf-32 or utf-16. on linux it's usually utf-32, and on windows it's usually (always?) utf-16. but you should not care about it. you see, in python, the unicode-strings are a separate data-type, and there's just no way to take a bytestring, and tell python: "from now on, you are an unicode-string, because i know that you are encoded in utf-16." the way it works is that you take a bytestring, and ask python to convert it into an unicode-string (and you also have to tell python the bytestring's charset). so while it might be, that the conversion from utf-16-bytestrings to unicode is sometimes faster thatn converting from utf-8-bytestrings to unicode, you can't be sure, because as i wrote above, the internal unicode-encoding is not fixed. > whereas IO and the databases mostly > speak UTF-8 - > so no, you can't dump it over the wire. > We Rubyists are a tad happier > because we now > have all in UTF-8 you mean that regexes, and all the methods of the string-class now are unicode-aware in ruby? :) gabor --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
ak wrote: > Ticket http://code.djangoproject.com/ticket/952 contain a complete > solution of this problem and I don't know why it was not merged into > the code but at the moment it is not matter and here is the reason why: > Since newforms library was born and the decision about using unicode > for clean_data was made, all these patches became unnecessary Not at all. Anton, read my summary that I posted as a reply to Michael first post. Specifying database encoding and keeping internals in unicode are two separate issues. #952 is still necessary but not enough to fix your bug. > because > now developers must use only unicode everywhere (templates, db etc) Actually the shouldn't :-). Newforms is now the only part of Django that works with unicode. I/O with th web (requests and templates) are now hotfixed to work with it in a way. Databases aren't. > or > manually recode all forms based on newforms from unicode to native > encoding and back. Ofcourse this is stupid May be it is. But it's a temporary inconvenience of newforms. Later database backend should do this automatically by using either 'utf-8' or DATABASE_CHARSET as I described in that my message. BTW, there were ideas here about really really forcing users to migrate all data into unicode/utf-8 and be the first guy on the block that would lead the trend. This is noble but hard and if I remember correctly this was decided against... > So, for me the quesion sounds like this: either newforms don't use > unicode to store clean_data and we can keep using 'legacy' character > sets, or django needs to drop all charsets support except of unicode. > Or it should convert strings back and forth everywhere LOL Incidentally you last 'LOL' is the option that Django have chosen :-). I'll try to explain. 'Unicode' is not a charset, or, more specifically, it is not represented with bytes. Python's native unicode string represent unicode characters in some internal format that just can't be dumped over the wire, be it to database or to the web. Because of this if Django would work internally in unicode it must encode everything it writes and decode everything it reads from outside. Converting from unicode to utf-8 is also encoding, and it does not happen automatically. When you say that db backend supports 'unicode' it actually means that db library under Django backend does the encoding itself. But whether it's done in the library or in Django backend we still need a setting for charset. Two settings actually: for the web (that we already have) and for db (that is implemented in #952). --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Hi, here's a summary what the different tickets are about: # 952 adds a database client encoding setting, DATABASE_CLIENT_CHARSET, for mysql and postgresql backends. For mysql, it uses the given charset in 'SET NAMES' to build the connection, except for mysql < 4.1. For postgresql, it does a 'SET CLIENT_ENCODING TO'. # 1356 sets the charset attribute of the mysql backend connection to 'utf8' for mysql version >= 4.1 # 3370 starts by explaining a traceback within newforms when you use utf8-encoded values with a form created by form_for_instance and has a patch that adds 'charset':'utf8' to the kwargs used in Database.connect() within DatabaseWrapper.cursor() Michael Radziej -- noris network AG - Deutschherrnstraße 15-19 - D-90429 Nürnberg - Tel +49-911-9352-0 - Fax +49-911-9352-100 http://www.noris.de - The IT-Outsourcing Company --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Guys The problem is simple but it was born a very long time ago. For MySQL 4.1 and higher there is hardcoded in django/db/backends/mysql/base.py: cursor.execute("SET NAMES 'utf8'") there were lots of tickets and messages in django-users complaining to this but in fact they all were ignored. Personally my company used to use patched django installation where this line was replaced to: cursor.execute("SET NAMES 'cp1251'") because all our templates were (and still are in the production environment) in windows-1251 encoding so we have had to use cp1251 to deal with db. Ticket http://code.djangoproject.com/ticket/952 contain a complete solution of this problem and I don't know why it was not merged into the code but at the moment it is not matter and here is the reason why: Since newforms library was born and the decision about using unicode for clean_data was made, all these patches became unnecessary because now developers must use only unicode everywhere (templates, db etc) or manually recode all forms based on newforms from unicode to native encoding and back. Ofcourse this is stupid and noone will do it because it's easier to migrate to utf-8 and forget about the problem. So, for me the quesion sounds like this: either newforms don't use unicode to store clean_data and we can keep using 'legacy' character sets, or django needs to drop all charsets support except of unicode. Or it should convert strings back and forth everywhere LOL Any other opinions ? --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---
Re: unicode issues in multiple tickets (#952, #1356, #3370) and thread about Euro sign in django-users
Michael Radziej wrote: > Hi, > > we have a bit of chaos here ... Tickets 3370, 1356 and probably 952 > all are about this problem, all are accepted, and #3370 and #1356 > have very similar patches. I ask everybody to continue discussion > here in django-developers, and I ask the authors of these three > tickets to work together to find out how to proceed. Right :-). I'll generalize my comment in #3370 here. There are, in fact, two separate issues. 1. First one (that #952 was intended to fix) is that we don't have a notion of a database internal encoding at all. This is bad because DB is as external to Django as the web and it can be in any encoding. Then there are two ways of dealing with it: - let Django encode data into a charset that a database expects - tell a database which encoding Django uses and let it to encode data into its internals #952 is implemented as a second variant and it looks like it works (in fact author of it is Julian Tarkhanov -- a well known unicode expert and advocate in russian blogosphere.. just giving credits :-) ) We really should have this thing regardless of Django's unicode or byte-string internals. 2. The second issue is an automatic conversion of unicode data for db backends that don't understand unicode. It's become relevant recently because people started to use newforms. If we accept #952 as it is then this should be resolved be encoding things into 'utf-8' inside backends. If we chose to reimplement database encoding support on django side then backend should encode into whatever encoding is stored in DATABASE_CHARSET setting. This is what things are like now. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Django developers" group. To post to this group, send email to django-developers@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-developers?hl=en -~--~~~~--~~--~--~---