[doc] fix doc and tutorials about hooks and operations

[optim] cache rset column structures When iterating over large resultsets (lot of colcumns, lot of lines) to build entities, caching column structure (which attribute or relation is stored in each column) can improve performance noticeably.

[js] make _loadAjaxHtmlHead aware of mod_concat-like urls _loadAjaxHtmlHead keeps track of already imported resources (js / css) to make sure they're not included twice, but with mod_concat-like urls, things get a bit more complicated : if BASE_URL/data??foo.js,bar.js is already imported, then loading : BASE_URL/data??bar.js,spam.js should only load spam.js

[etwist] in debug-mode, concat-file is rebuilt if one of the concatenated file is more recent

[facets] display splash image while facets are reloading content (closes #1624851)

[facets] mark titles of active facets with red dot (closes #1624857)

[facet] add missing selector to HasRelationFacet

[facets] proper fix for #1625112: we should rather call unicode() as we always manipulate string as unicode internally

merge

[etwist] Generates apache's mod_concat-like urls for js and css Uses the lists of js, css and ie_css included in order to generate urls compatible with Apache mod_concat (ie: http://test.fr/??file1.js,file2.js). Css files may use `media` option or may be for IE only ; this patch does not concatenate css filenames for different media types or different IE specs. It concatenates only files served under cubicweb's `data` urls and keeps original include order. The generated urls are added to the HTML header as usual. HOW-TO: - calling HTMLHead(datadir_url) will enable the url concatenation - calling HTMLHead() will not benefit of url concatenation IMPORTANT NOTE: - fckeditor can no more be served under `/data/` segment

handle js/ css file concatenation with twisted

put data version md5 into a new url segment to ease static serving

add cwctl command to generate a consolidated data directory with all cubes/ cubicweb data directories content to be delivered statically

[etwist] refactor child resources management

[vreg] vregistry._select_best was needlessly instanciating NoSelectableObject (closes #1626708) in large selections, this incurred a non negligible cost (add_relations with 2000 inlined relations runs 7% faster with this patch)

backport stable

merge stable

merge oldstable

backport oldstable

[page data] don't call unload_page_data too many time (closes #1518398)

[request] don't init tabindex unnecessarily, since it's added to page data which will makes an ajax call to unload_page_data necessary

[deprecation] kill a remaining 3.6 warning

[web config] fix md5 version to ensure cubes order and to include cubicweb version

merged stable bugfix

merged oldstable bugfix

place the disable hooks context manager at the right place (closes: #1626694) hooks control is reset at each commit so we need to have a smaller scope for the CM in the 3.10.9 migration script

backport stable into default

backport oldstable into stable

[js] fix inArray usage in reorderTabIndex() : inArray returns -1 if element is not found, not 'false'

[rql] basic support for regexp-based pattern matching

[rql] support CAST() Function

stable is now 3.12

oldstable is now 3.11

[js] delete empty js files (incluced in cw.ajax.js)

[sql generation] fix crash on outer join + ambiguous inlined relation + attribute selection

[rql rewrite] backout 7177:f2a976cf7dac : since EXISTS is the variable scope, we want to insert snippets here anyway (other solution would be to NOT insert snippets at all in such case). Closes #1625464

[book] fix documentation + add chpasswd.py script

[facets] fix unicode encode error; closes #1625112

[facets] fix RangeFacet get_widget: do no return a widget unless there are at least 2 values; closes #1625203

merge

Backed out changeset 7a58581d842c This changeset breaks a number of CW tests, which depends on pre add hooks. I'm backing it out, so that we can start the need-for-speed sprint with green tests (and possibly rework this changeset)

small reformatting (but test still fails)

fix tests broken by fcb8932082a5 (continued)

fix bad handling of inlined relations when entities of different etypes are passed as subject to add_relations

fix failing test_migration test caused by schema change

fix tests broken by fcb8932082a5

merge back stable

fix server/test/unittest_checkintegrity (failing since 70538ea2532d)

fix implementation of repository.glob_add_relations (closes ##1625257) the internal data structure had changed during implementation and the code had not been updated...

[server] improve the speed of setting relations between entities (closes #1625257) The main idea is to add methods equivalent to session.add_relation and repository.glob_add_relation which handle several relations in one call. Speed gain results from: * using cursor.executemany to run SQL statements * factorizing some code which otherwise has to be performed for each relation, such as context manager creation before calling hooks or to enable security, creation of the EditedEntity instance (when several inlined relations are set on a single entity, and consequently when refreshing the cached entity) benchmark runs 1.1x faster for non inlined entities and 125x faster for inlined entities :-)

added test case to monitor speed improvements in the repository code. To use it, add a test in a patch, then hack CW to speed it up in another patch, updating the test if needed. By running the test with and without the second patch applied you can easily see the speed improvements

backport stable changes to default

[js] update jquery.corner.js to the latest version (uses native css support for browser that provide it); closes #1625065

[js/css] remove old versions of ui.tabs/slider etc.; closes #1625085

[devtools] make return value of get_response_header(raw=True) consistent wrt get_header(raw=True) i.e. get_response_header(raw=True) now returns the first element of the list rather than the list itself.

[widgets] handle no value label for tristate radios, put default separator on class for easier redefinition - closes #1624839

[doc] update of the update

[tz postgres support] we've to commit the connection once time-zone is set, else we may loose the setting

[tz support] more tests

[dataimport] actually commit every desired number...

[doc] major update of the install documentation

[facets] make rangefacet comply with start_unfolded (closes #1615869)

[testlib] make a clear distinction between input / output HTTP headers The get_header() / set_header() API is the same as for CubicWebRequestBase : - get_header() returns the **request** header - set_header() / add_header() works on the **response** header FakeRequest adds two additional methods : - get_response_header() returns the **reponse** header - set_request_header() adds a **request** header FakeRequest now uses the `headers_out` ineherited from CubicWebRequestBase. It also uses a `_headers_in` attribute

[entity] create_entity/cw_instantiate should initialize entity's cache to avoid later queries

[testlib] refactor create_user and grant_permission to make them usable from pre_setup_database

[3.12 regression] don't access to __class__ directly, it may raise AttributeError

[testlib] fix typos / NameErrors in devtools/__init__.py

[schema serialization] add warning on missing group

backport stable

[testlib] fix bug in backup file name generation: it may erroneously replace '-' by '_' in database **parent directories**. This should fix random error when test are run under apycot.

backport stable

[schema diagram] proper url generation (closes #1621276)

[testlib] fix readlb config handling, avoid systematic recreation of db with postgres (closes #1618089)

[entities] fix cw_instantiate typo

[entities/test] add failing test for cw_instantiate

[timezone] ensure tzinfo is None

[entity] use a list comprehension

Added tag cubicweb-debian-version-3.12.2-1 for changeset 6d418fb3ffed

Added tag cubicweb-version-3.12.2 for changeset 5a9b6bc56538

[packaging] prepare the release of version 3.12.2

merge changes from stable branch

[facet] ignore range extremas to avoid filtering out rows where value is NULL (closes #1251192)

Backed out changeset e0e802afa8e1, widgets are actually defined below

[facet] refactor: a simple class attribute is enough

[native source] eid_creation_lock actually lock the eid connection creation and usage

[dataimport] allow to specify columns on which result should be sorted in lazydbtable

[entity creation] ensure this is a relation and not an attribute before moving things into pending_relations because value is a list Using a list for attributes may be used for later type specific casting (e.g. timeseries)

[fckeditor] fix fckeditor relative url: use baseuri()

Added tag cubicweb-debian-version-3.12.1-1 for changeset 07e2c9c7df26

Added tag cubicweb-version-3.12.1 for changeset 59701627adba

3.12.1

i18n update

backport stable

[sql generation] fix missing join with outer joins w/ rhs invariant variable... ... where the principal where not the processed relation. This led to major refactoring in outer join handling, which is now much clearer and powerful. Closes #1579228

[test, cw props] ensure we can't add user specific / system-wide properties we had this pb on an historical instance in our intranet (eg. running for about 8 years).

[ms planner] fix buggy temp table name generation, may generate duplicated unique id with subqueries/union