2009-02-08T03:18:49 <ThomasWaldmann> TheSheep: why is anchor setting in creole's image_emit?
2009-02-08T04:13:31 <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 4281:050428d1c044 1.8/MoinMoin/ (4 files in 4 dirs): (log message trimmed)
2009-02-08T04:13:31 <CIA-38> Fix processing of anchors (fragment identifiers), see details below
2009-02-08T04:13:31 <CIA-38> Fixed creole and wiki parser, other parsers might need similar fixes.
2009-02-08T04:13:31 <CIA-38> IDs that are valid fragment identifiers should see no changes.
2009-02-08T04:13:31 <CIA-38> IDs with blanks, non-ASCII chars etc. are now sanitized in the same way for
2009-02-08T04:13:35 <CIA-38> links as well as for link targets, so the user editing a page won't have to
2009-02-08T04:13:37 <CIA-38> bother with it. [[#foo bar]] will link to <<Anchor(foo bar)>> or = foo bar =.
2009-02-08T12:46:23 <ThomasWaldmann> moin
2009-02-08T14:39:36 <waldi> http://localhost:8080/%3CMoinMoin.util.iri.Iri%20object%20at%200x9f4be0c%3E
2009-02-08T14:39:44 <waldi> ups
2009-02-08T14:43:19 <waldi> okay, first i need the possibility to use + on iris
2009-02-08T15:08:11 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4888:db5f7e522149 1.8-dom-bblank/MoinMoin/util/_tests/test_iri.py: Add tests for IRI
2009-02-08T15:08:13 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4889:9994b01d2677 1.8-dom-bblank/MoinMoin/util/iri.py: Add generic IRI class
2009-02-08T15:08:14 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4890:d9858936a875 1.8-dom-bblank/MoinMoin/util/iri.py: IRI - Add quote and unquote functions
2009-02-08T15:08:15 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4891:844dc2913979 1.8-dom-bblank/MoinMoin/ (8 files in 4 dirs): Page, Converter, Compatibility formatter - Use Iri
2009-02-08T15:08:17 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4892:9336ecbf6e05 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Factor out parsing code
2009-02-08T15:08:19 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4893:897d78eda06f 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Handle unquote in one run
2009-02-08T15:08:22 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4894:51a588fe3dd3 1.8-dom-bblank/.hgtags: Fix tags
2009-02-08T16:07:25 * ThomasWaldmann looks at the iri stuff
2009-02-08T16:09:52 <waldi> hmm, py.test does not support properties:
2009-02-08T16:09:53 <waldi> assert <MoinMoin.util.iri.Iri object at 0x8cc5fec>.query_quoted == 'Neu%25?'
2009-02-08T16:18:20 <ThomasWaldmann> waldi: is it intended that line 137 does not use "is not None"?
2009-02-08T16:19:16 <waldi> does not matter
2009-02-08T16:19:40 <waldi> empty schema are not allowed, while other parts can be either empty or missing
2009-02-08T16:20:17 <ThomasWaldmann> ok, but better use unicode string constant in __unicode__
2009-02-08T16:20:50 <waldi> yeah
2009-02-08T16:24:38 <dreimark> moin
2009-02-08T16:26:54 <ThomasWaldmann> waldi: for what is the _q stuff in _parse?
2009-02-08T16:28:15 <waldi> ThomasWaldmann: unused for now
2009-02-08T16:30:11 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4895:526bc8282148 1.8-dom-bblank/MoinMoin/util/ (_tests/test_iri.py iri.py): Iri - Add quoted forms of query
2009-02-08T16:30:12 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4896:33020976f6c1 1.8-dom-bblank/MoinMoin/util/ (_tests/test_iri.py iri.py): Iri - Add quoted forms of fragment
2009-02-08T16:30:13 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4897:2a490fe7f9e9 1.8-dom-bblank/MoinMoin/util/_tests/test_iri.py: Tests for Iri - Add tests for full quoted forms
2009-02-08T16:34:34 <ThomasWaldmann> waldi: quote_*_rules = (...) + quote_rules # usually faster
2009-02-08T16:36:30 <dreimark> TheSheep: last test of testAnchor fails currently in 1.8
2009-02-08T16:38:34 * ThomasWaldmann runs tests
2009-02-08T16:38:54 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4898:781c1c776d53 1.8-dom-bblank/MoinMoin/util/ (_tests/test_iri.py iri.py): Iri - Add quoted forms of path
2009-02-08T16:38:55 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4899:15a08b45290e 1.8-dom-bblank/MoinMoin/util/ (_tests/test_iri.py iri.py): Iri - Add quoted forms of authority
2009-02-08T16:38:57 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4900:ced69b7e383c 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Use always unicode strings
2009-02-08T16:39:31 <waldi> ThomasWaldmann: irrelevant compared to the re setup costs
2009-02-08T16:41:52 <waldi> hmm, is there anything which can't be successfully parsed as uri/iri? (at least if you ignore the character quoting rules)
2009-02-08T16:52:10 <waldi> TypeError: 'NoneType' object is unsubscriptable
2009-02-08T16:52:11 <waldi> yeah
2009-02-08T16:52:53 <ThomasWaldmann> waldi: re is setup only once, but you often iterated over those lists of tuples
2009-02-08T16:53:50 <ThomasWaldmann> thus, the usually ASCII stuff should be rather at the beginning not after all those unicode ranges
2009-02-08T16:54:18 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4901:fefa7e1a88bb 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Finish support for quoted forms
2009-02-08T16:54:19 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4902:3dc980de0557 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Error out if input can't be parsed as IRI
2009-02-08T16:54:20 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4903:bbdd3d2974b7 1.8-dom-bblank/MoinMoin/converter2/link.py: Link converter - Use Iri correctly
2009-02-08T16:54:26 <waldi> ThomasWaldmann: ah, yes
2009-02-08T16:58:06 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4904:c1c3cdd5e953 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Move the unicode ranges at the end of the match rules
2009-02-08T17:06:15 <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 4282:57ab9fdceac0 1.8/MoinMoin/parser/_tests/test_text_creole.py: fix creole parser's anchor test
2009-02-08T17:09:15 <ThomasWaldmann> waldi: 158
2009-02-08T17:09:35 <ThomasWaldmann> u''.join(...
2009-02-08T17:09:46 <waldi> yep
2009-02-08T17:11:21 <waldi> oh, error in the logic
2009-02-08T17:14:38 <ThomasWaldmann> that property stuff seems weird
2009-02-08T17:14:54 <ThomasWaldmann> sometimes gets 1 value, sometimes a tuple
2009-02-08T17:16:16 <ThomasWaldmann> waldi: some docstrings and tests would be nice :)
2009-02-08T17:16:29 <waldi> ThomasWaldmann: no, the properties always uses one value to the outside. the internal value is saved as tuple
2009-02-08T17:16:32 <waldi> yes
2009-02-08T17:17:17 <ThomasWaldmann> ah, ok
2009-02-08T17:21:36 <ThomasWaldmann> waldi: _overall_re could move directly below overall_rules and become overall_re, so one can use it externally
2009-02-08T17:22:02 <ThomasWaldmann> (and maybe rename to iri_re or so)
2009-02-08T17:22:45 <ThomasWaldmann> if you use match, the ^ at the beginning is not needed, right?
2009-02-08T17:23:03 <ThomasWaldmann> (without that, other code could use it for re.search maybe)
2009-02-08T17:24:40 <waldi> this re can't be used to find uris. every string is a valid (relative) uri
2009-02-08T17:30:56 <ThomasWaldmann> hmm, right, the old problem...
2009-02-08T17:32:23 <ThomasWaldmann> waldi: for practical purposes, the authority quoting/coding could be rather punicode
2009-02-08T17:35:02 <ThomasWaldmann> what we also need is a "interpret this iri relative to another iri"
2009-02-08T17:36:34 <waldi> ThomasWaldmann: no, punicode is only defined for URI. IRI explicitely uses the decoded form
2009-02-08T17:42:13 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4905:946fb6facf99 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Don't requote strings, factor out fullquoted support
2009-02-08T17:42:15 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4906:4a500f6ef235 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Factor out quoted support
2009-02-08T17:42:16 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4907:2197de3d52c7 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Add docstrings
2009-02-08T17:42:27 <ThomasWaldmann> i mean that Iri class offers a transformation to uri (by utf-8 encoding, urlencoding)
2009-02-08T17:43:15 <ThomasWaldmann> it is just not practically useful in the domain/server part as there won't be a schn%XX%XXrle.com domain
2009-02-08T17:43:30 <ThomasWaldmann> but rather a xn--schnrle-foobar.com
2009-02-08T17:44:53 <waldi> do we want to encode anything to URIs in the html output should we assume that the current browsers works with IRI also?
2009-02-08T17:45:05 <waldi> only for http we need to convert them
2009-02-08T17:47:55 <ThomasWaldmann> i don't know if any current browser can use IRIs directly
2009-02-08T17:49:45 <waldi> <a class="nonexistent" href="http://localhost:8080//äöü">äöü</a>
2009-02-08T17:49:46 <waldi> lets check
2009-02-08T17:51:17 <waldi> so gecko and w3m supports it at least. but there is ony straight /
2009-02-08T17:51:52 <ThomasWaldmann> try to access schnürle.com :)
2009-02-08T17:52:49 <waldi> <a href="http://schnürle.com/">http://schnürle.com/</a>
2009-02-08T17:53:04 <waldi> correctly expands to http://xn--schnrle-q2a.com/
2009-02-08T17:56:01 <ThomasWaldmann> ok, so some browsers use IRIs internally. Likely except older ones and always-broken IE.
2009-02-08T17:56:22 <waldi> yeah
2009-02-08T17:57:58 <ThomasWaldmann> so that means that if we want to support those, we need to transform to URIs ourselves
2009-02-08T18:05:13 <ThomasWaldmann> that quoting stuff is confusing and those docstrings don't help much as they are not telling more than the method name
2009-02-08T18:05:56 <ThomasWaldmann> one should simply tell when to use what
2009-02-08T18:06:25 <TheSheep> ThomasWaldmann: werkzeug has a 'fix url' function
2009-02-08T18:06:34 <TheSheep> ThomasWaldmann: it does some magic
2009-02-08T18:06:49 <TheSheep> it's for user-entered urls mainly
2009-02-08T18:08:20 <ThomasWaldmann> ah, nice
2009-02-08T18:08:32 <ThomasWaldmann> TheSheep: did you look at my anchor changesets?
2009-02-08T18:09:06 * TheSheep looks
2009-02-08T18:12:00 <ThomasWaldmann> it is just action similarly on anchors as on pagenames. we allow unicode pagenames and just quote/encode them somehow in links as well as at the link target.
2009-02-08T18:12:12 <ThomasWaldmann> ...acting...
2009-02-08T18:13:14 <TheSheep> ThomasWaldmann: aren't they already encoded in the url?
2009-02-08T18:14:05 <ThomasWaldmann> the code often just used #%s % anchor, sometimes it used url_quote_plus, sometimes it used sanitize_to_id ...
2009-02-08T18:14:41 <ThomasWaldmann> it try to make it consistent
2009-02-08T18:15:30 <ThomasWaldmann> as long as someone specifies valid IDs / fragment names as per html4 rfc, the sanitize transformation should not do any change
2009-02-08T18:16:56 <ThomasWaldmann> but you can now also have [[#hä hä!]] working (same for <<Anchor(...)>> and {{#...}}). it won't be a pretty ID, but consistent.
2009-02-08T18:17:20 <TheSheep> cool
2009-02-08T18:17:46 <ThomasWaldmann> also, you likely can refer now to headlinges by just repeating their text in the anchor link
2009-02-08T18:18:18 <TheSheep> that's nice, no need to manually retype them
2009-02-08T18:18:26 * ThomasWaldmann .oO(my wetware output processor is a bit sloppy currently)
2009-02-08T18:18:52 <TheSheep> mine is almost inoperational
2009-02-08T18:20:26 <ThomasWaldmann> as a side effect of those fixes, some content that was manually made working before now needs removing the workarounds
2009-02-08T18:21:11 <ThomasWaldmann> like using [[#foobar]] to link to <<Anchor(foo_bar)>>
2009-02-08T18:30:02 * ThomasWaldmann updates CHANGES
2009-02-08T18:35:27 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4908:f9fe47c6a8f8 1.8-dom-bblank/MoinMoin/converter2/link.py: Link converter - Use Iri correctly
2009-02-08T18:35:29 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4909:6628a2d72c16 1.8-dom-bblank/MoinMoin/util/iri.py: Iri - Add more documentation
2009-02-08T18:42:08 <CIA-38> Bastian Blank <bblank@thinkmo.de> default * 4910:a6921d3f40a6 1.8-dom-bblank/MoinMoin/converter2/link.py: Link converter - Use root URL as complete URL
2009-02-08T18:52:35 <dreimark> hmm SlideShow from 1.7-extension looks strange if you look at one slide
2009-02-08T18:53:18 <dreimark> the links are listed one per row
2009-02-08T19:01:21 <ThomasWaldmann> css?
2009-02-08T19:04:41 <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 4283:2f9b78965bce 1.8/docs/CHANGES: updated CHANGES
2009-02-08T19:14:44 <ThomasWaldmann> <<Anchor(Paramétrage)>>
2009-02-08T19:14:44 <ThomasWaldmann> == Paramétrage ==
2009-02-08T19:14:52 <ThomasWaldmann> no need for such stuff any more
2009-02-08T19:17:09 <dreimark> ThomasWaldmann: css is enabeld I get the blue colored headings on the page itselfs
2009-02-08T19:18:41 <ThomasWaldmann> does it load slideshow.css?
2009-02-08T19:20:06 <dreimark> ThomasWaldmann: fixed. wrong css place
2009-02-08T19:21:50 * ThomasWaldmann runs make dist
2009-02-08T19:30:16 <ThomasWaldmann> http://static.moinmo.in/files/moin-1.8.2.tar.gz < please help testing
2009-02-08T19:34:05 <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 4284:1f0db10c207f 1.8/docs/ (INSTALL.html UPDATE.html): make install-docs interwiki
2009-02-08T19:36:31 <dreimark> download startet
2009-02-08T19:57:25 <ThomasWaldmann> btw, if someone has a konqueror, safari, opera, IE, a test of the gui editor with those would be nice
2009-02-08T19:57:51 <dreimark> can check konqueror
2009-02-08T20:00:54 <dreimark> Version 4.1.3 (KDE 4.1.3) does not show the gui button
2009-02-08T20:02:21 <ThomasWaldmann> check on fckeditor.net if it works with the demo there
2009-02-08T20:03:58 <dreimark> it doesn't
2009-02-08T20:06:29 <dreimark> ie6 for linux shows the gui editor and I can edit using it. saving needs a reload to get it shown. I've no idea if that is similiar on a win box.
2009-02-08T20:07:24 <ThomasWaldmann> hmm, they don't explicitely state compatibility to konqueror, so maybe it is ok like it is
2009-02-08T20:09:29 <dreimark> konq 3.5.7 also does not work
2009-02-08T20:09:49 * dreimark looks on opera
2009-02-08T20:11:26 <dreimark> opera works too
2009-02-08T20:16:37 * ThomasWaldmann has the finger on the trigger :)
2009-02-08T20:28:54 <ThomasWaldmann> dreimark: (or anyone else): anything unusual so far?
2009-02-08T20:29:27 <dreimark> nothing yet, looks good
2009-02-08T20:29:58 <ThomasWaldmann> ok, then let's release it :)
2009-02-08T20:38:06 <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 4285:6a287955a262 1.8/.hgtags: tagged release 1.8.2
2009-02-08T21:16:03 <waldi> ThomasWaldmann: hmm, how do the browsers react on % in hostnames of the uris? from the rfc this is allowed and the useragent should do the punycode conversion on its own
2009-02-08T21:20:33 <ThomasWaldmann> oh really? which rfc?
2009-02-08T21:20:48 <waldi> 3986
2009-02-08T21:21:06 <waldi> reg-name is the definition
2009-02-08T21:21:17 <ThomasWaldmann> that's also rather new, so one can't be sure everyone implements it that way, right?
2009-02-08T21:24:01 <waldi> 4 years. and even the old URI rfc lists it this way
2009-02-08T21:24:29 <ThomasWaldmann> ok, so if you are sure(tm) ... :)
2009-02-08T21:25:32 <waldi> the reason that punycode is often used expanded was the slow adaption in the user agents, not forced by the standards
2009-02-08T21:26:47 <ThomasWaldmann> btw, maybe fetch some comments from #python(.de), I think that stuff is generally interesting
2009-02-08T21:28:35 <ThomasWaldmann> maybe add some more comments about motivation and usage first :)
2009-02-08T21:28:50 <ThomasWaldmann> not everyone will happily read the rfc
2009-02-08T21:34:10 <ThomasWaldmann> btw, for what is that ^ at the beginning of the re needed? it also has no $ at the end. And it is not needed for re.match anyway, so?
2009-02-08T21:36:17 <waldi> no, it is not
2009-02-08T21:38:37 <ThomasWaldmann> maybe add some ":" after the attr names and before the description in the docstrings. maybe not required but looks better.
2009-02-08T21:40:37 <ThomasWaldmann> as this stuff is rather generic and not related to MoinMoin, maybe remove references to MoinMoin
2009-02-08T21:41:13 <ThomasWaldmann> maybe another license also could make sense
2009-02-08T21:44:48 <ThomasWaldmann> like the python license for example, so other python projects could also use this
2009-02-08T21:51:35 <ThomasWaldmann> waldi: btw, in __init__ all args should go through unicode() to make it fail early (and the cast from _parse maybe could move there also)
2009-02-08T21:57:35 <ThomasWaldmann> waldi: all ord() arguments could be u"x". for ascii not required (result is same after ord()), but it would point out that we are dealing with unicode code points.
2009-02-08T21:57:54 <ThomasWaldmann> maybe also makes 2to3 happier
2009-02-08T21:59:45 <ThomasWaldmann> waldi: there are some codepoints that require ucs4 python, maybe that should be noted or asserted or handled somehow
2009-02-08T22:03:47 <ThomasWaldmann> waldi: s/__get_all_/__get_/g
2009-02-08T22:15:05 <waldi> the large codepoints will bail out anyway if a conversion is tried. i use ord to get them from an existing string
2009-02-08T22:15:55 <waldi> okay, if we want to do punycode conversion somewhere, the authority part needs to be split in the three parts userinfo, host and port
2009-02-08T22:23:22 <waldi> i think i split that into several classes, it gets too large MoinMoin: MoinMoinChat/Logs/moin-dev/2009-02-08 (last edited 2009-02-08 02:30:02 by IrcLogImporter)
