This is a rather large refactoring. The "lookup traversal" code was
splitted out from the setup_joins. There is now names_to_path() method
which does the lookup traveling, the actual work of setup_joins() is
calling names_to_path() and then adding the joins found into the query.
As a side effect it was possible to remove the "process_extra"
functionality used by genric relations. This never worked for left
joins. Now the extra restriction is appended directly to the join
condition instead of the where clause.
To generate the extra condition we need to have the join field
available in the compiler. This has the side-effect that we need more
ugly code in Query.__getstate__ and __setstate__ as Field objects
aren't pickleable.
The join trimming code got a big change - now we trim all direct joins
and never trim reverse joins. This also fixes the problem in #10790
which was join trimming in null filter cases.
The trim argument was used by split_exclude() only to trim the last
join from the given lookup. It is cleaner to just trim the last part
from the lookup in split_exclude() directly so that there is no need
to burden add_filter() with the logic needed for only split_exclude().
F() expressions reuse joins like any lookup in a .filter() call -
reuse multijoins generated in the same .filter() call else generate
new joins. Also, lookups can now reuse joins generated by F().
This change is backwards incompatible, but it is required to prevent
dict randomization from generating different queries depending on
.filter() kwarg ordering. The new way is also more consistent in how
joins are reused.
The select_related code got confused when it needed to travel a
reverse relation to a model which had different parent than the
originally travelled relation.
Thanks to Trac aliases shauncutts for report and ungenio for original
patch (committed patch is somewhat modified version of that).
The problem is the same as in #10888 which was reintroduced when
bulk_insert was added. Thanks to Jani Tiainen for report, patch and
also testing the final patch on Oracle GIS.
The dupe avoidance logic was removed as it doesn't seem to do anything,
it is complicated, and it has nearly zero documentation.
The removal of dupe_avoidance allowed for refactoring of both the
implementation and signature of Query.join(). This refactoring cascades
again to some other parts. The most significant of them is the changes
in qs.combine(), and compiler.select_related_descent().
The Query.select and Query.select_fields were collapsed into one list
because the attributes had to be always in sync. Now that they are in
one attribute it is impossible to edit them out of sync.
Similar collapse was done for Query.related_select_cols and
Query.related_select_fields.
There was a bug introduced in #18676 which caused fast-path deletes
implemented as "DELETE WHERE pk IN <subquery>" to fail if the SELECT
clause contained additional stuff (for example extra() and annotate()).
Thanks to Trac alias pressureman for spotting this regression.
RETURNING is an extension of the SQL standard, which is not implemented
the same by all databases. Allow DatabaseOperations.return_insert_id to
return a None to allow for other 3rd party backends with a different
implementation.
Objects can be fast-path deleted if there are no signals, and there are
no further cascades. If fast-path is taken, the objects do not need to
be loaded into memory before deletion.
Thanks to Jeremy Dunck, Simon Charette and Alex Gaynor for reviewing
the patch.
When doing deeper than one level select_related() + only queries(), the
code introduced in b6c356b7bb errored
incorrectly.
Thanks to mrmachine for report & test case.
In an ideal world, nothing except django.db.models.query should have to
import stuff from django.models.sql.*. A few things were needing to get
hold of sql.constants.LOOKUP_SEP, so this commit moves it up to
django.db.models.constants.LOOKUP_SEP.
There are still a couple of places (admin) poking into sql.* to get
QUERY_TERMS, which is unfortunate, but a slightly different issue and
harder to adjust.
The joins for nested nullable foreign keys were often created as INNER
when they should have been OUTER joins. The reason was that only the
first join in the chain was promoted correctly. There were also issues
with select_related etc.
The basic structure for this problem was:
A -[nullable]-> B -[nonnull]-> C
And the basic problem was that the A->B join was correctly LOUTER,
the B->C join not.
The major change taken in this patch is that now if we promote a join
A->B, we will automatically promote joins B->X for all X in the query.
Also, we now make sure there aren't ever join chains like:
a LOUTER b INNER c
If the a -> b needs to be LOUTER, then the INNER at the end of the
chain will cancel the LOUTER join and we have a broken query.
Sebastian reported this problem and did also major portions of the
patch.
The ORM generated a query with INNER JOIN instead of LEFT OUTER JOIN
in a somewhat complicated case. The main issue was that there was a
chain of nullable FK -> non-nullble FK, and the join promotion logic
didn't see the need to promote the non-nullable FK even if the
previous nullable FK was already promoted to LOUTER JOIN. This resulted
in a query like a LOUTER b INNER c, which incorrectly prunes results.
* Renamed smart_unicode to smart_text (but kept the old name under
Python 2 for backwards compatibility).
* Renamed smart_str to smart_bytes.
* Re-introduced smart_str as an alias for smart_text under Python 3
and smart_bytes under Python 2 (which is backwards compatible).
Thus smart_str always returns a str objects.
* Used the new smart_str in a few places where both Python 2 and 3
want a str.
Cleared aggregations on add_date_select method so only distinct dates
are returned when dealing with a QuerySet that contained aggregations.
That would cause the query set to return repeated dates because it
would look for distinct (date kind, aggregation) pairs.
At least Oracle needs parentheses in negated where conditions, even if
there is only single condition negated. Fixed this by reverting to old
logic in that part of as_sql() and adding a comment about this.
I did not investigate why the parentheses are needed. The original
offending commit was bd283aa844.
Made sure the WhereNode.as_sql() handles various EmptyResultSet and
FullResultSet conditions correctly. Also, got rid of the FullResultSet
exception class. It is now represented by '', [] return value in the
as_sql() methods.
This commit tackles a couple of issues. First, in certain cases there
were some mixups if field.attname or field.name should be deferred.
Field.attname is now always used.
Another issue tackled is a case where field is both deferred by
.only(), and selected by select_related. This case is now an error.
A lot of thanks to koniiiik (Michal Petrucha) for the patch, and
to Andrei Antoukh for review.
When order_by causes new joins to be added to the query, the joins must
be LEFT OUTER joins for nullable relations, otherwise the order_by
could cause the results to be altered. This commit fixes the logic to
only promote new joins, previously all joins in the order_by lookup
path were promoted.
Thanks to Bruno Desthuilliers for spotting this corner case.
Databases with update_can_self_select = False (MySQL for example)
generated non-necessary queries when saving a multitable inherited
model, and when the save resulted in update.
Fixed#18248 -- proxy models were added to included_inherited_models
in sql.query.Query. The variable is meant to be used for multitable
inheritance only. This mistake caused problems in situations where
proxy model's query was reused.
Fixed#17957 -- when using Oracle and character fields, the fields
were set null = True to ease the handling of empty strings. This
caused problems when using multiple databases from different vendors,
or when the character field happened to be also a primary key.
The handling was changed so that NOT NULL is not emitted on Oracle
even if field.null = False, and field.null is not touched otherwise.
Thanks to bhuztez for the report, ramiro for triaging & comments,
ikelly for the patch and alex for reviewing.
QuerySet had previously some complex logic for dealing with nullable
fields in negated add_filter() calls. It seems the logic is leftover
from a time where the WhereNode wasn't as intelligent in handling
field__in=[] conditions.
Thanks to aaugustin for comments on the patch.
only consider some fields (PostgreSQL only).
For this, the ``distinct()`` QuerySet method now accepts an optional
list of model fields names and generates ``DISTINCT ON`` clauses on
these cases. Thanks Jeffrey Gelens and Anssi Kääriäinen for their work.
Fixes#6422.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@17244 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This is the old Query.as_sql() method revived: it's like Query.__str__,
but the parameters aren't substituted into the placeholders. Thus, it's
a more accurate representation of the SQL the (default) backend will
see. Entirely internal.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@16655 bcc190cf-cafb-0310-a4f2-bffc1f526a37
A number of people worked on this patch over the years -- Hawkeye, Colin Grady,
KBS, sakyamuni, anih, jdemoor, and Issak Kelly. Thanks to them all, and
apologies if I missed anyone.
Special thanks to Dan Fairs for picking it up again at the end and seeing this
through to commit.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@16058 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Turns out that a lot more than just SELECT can return data, and this list is
very hard to define up front in a cross-database manner. So let's just assume
that anyone using raw() is at least halfway competant and can deal with
the error messages if they don't use a data-returning query.
Thanks to Christophe Pettus for the patch.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@15803 bcc190cf-cafb-0310-a4f2-bffc1f526a37
as that can lead to incorrect SQL being generated for the query. Thanks to mat
for the report and test, tobias for the fix, and Alex for review.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@12830 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This monster of a patch is the result of Alex Gaynor's 2009 Google Summer of Code project.
Congratulations to Alex for a job well done.
Big thanks also go to:
* Justin Bronn for keeping GIS in line with the changes,
* Karen Tracey and Jani Tiainen for their help testing Oracle support
* Brett Hoerner, Jon Loyens, and Craig Kimmerer for their feedback.
* Malcolm Treddinick for his guidance during the GSoC submission process.
* Simon Willison for driving the original design process
* Cal Henderson for complaining about ponies he wanted.
... and everyone else too numerous to mention that helped to bring this feature into fruition.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@11952 bcc190cf-cafb-0310-a4f2-bffc1f526a37
See `docs/topics/db/raw.txt` for details.
Thanks to seanoc for getting the ball rolling, and to Russ for wrapping things up.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@11921 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This enables querysets with an extra clause to be used in an __in filter; as a side effect, it also means that as_sql() now returns the correct result for any query with an extra clause.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@10648 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Using select_related(...) across a nullable relation to a multi-table
model inheritance situation no longer excludes results. Thanks to AdamG
for a test demonstrating part of the problem.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@10136 bcc190cf-cafb-0310-a4f2-bffc1f526a37
In extreme cases, some fields are expensive to load from the database
(e.g. GIS fields requiring conversion, or large text fields). This
commit adds defer() and only() methods to querysets that allow the
caller to specify which fields should not be loaded unless they are
accessed.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@10090 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Large portions of this are needed for #5420, so I implemented it fully.
Thanks to Ryan Kelly for an initial patch to get this started.
Refs #5420.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@10083 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Ensure to read the documentation before blindly enabling this: requires some
code audits first, but might well be worth it for busy sites.
Thanks to nicferrier, iamseb and Richard Davies for help with this patch.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@10029 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This removes a long-standing FIXME in the update() handling and allows for
greater flexibility in the values passed in. In particular, it brings updates
into line with saves for django.contrib.gis fields, so fixed#10411.
Thanks to Justin Bronn and Russell Keith-Magee for help with this patch.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@10003 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This fixes a broad class of bugs involving filters that look for missing
related models and fields. Most of them don't seem to have been reported
(the added tests cover the root cause). The exception is that this has
also fixed#9868.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9979 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This includes a fairly large refactor of the update() query path (and
the initial portions of constructing the SQL for any query). The
previous code appears to have been only working more or less by accident
and was very fragile.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9967 bcc190cf-cafb-0310-a4f2-bffc1f526a37
If an update only affected an ancestor model (not the child), we were
returning 0 for the number of rows updated. This could have been
misleading if the value is used to detect an update occuring. So we now
return the rowcount from the first non-trivial query that is executed
(if any). Still a slight compromise, but better than what we had.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9966 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Some results were inadvertently being excluded if we were ordering across a
nullable relation which itself ordering by a non-nullable relation.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9916 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This involves a slight change to the interaction of annotate() and values() clauses that specify a list of columns. See the docs for details.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9888 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Many thanks to:
* Nicolas Lara, who worked on this feature during the 2008 Google Summer of Code.
* Alex Gaynor for his help debugging and fixing a number of issues.
* Malcolm Tredinnick for his invaluable review notes.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9792 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Many thanks to:
* Nicolas Lara, who worked on this feature during the 2008 Google Summer of Code.
* Alex Gaynor for his help debugging and fixing a number of issues.
* Justin Bronn for his help integrating with contrib.gis.
* Karen Tracey for her help with cross-platform testing.
* Ian Kelly for his help testing and fixing Oracle support.
* Malcolm Tredinnick for his invaluable review notes.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9742 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This extends previous functionality that allowed passing Query objects as the
rvals to filters. You can now pass QuerySets, which requires less poking at
opaque attributes. See the documentation of the "__in" lookup type for the
details.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9701 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Extricated the code that works directly with SQL columns (standard
"where" stuff) from the the code that takes SQL fragments and combines
it with lookup types and values. The latter portion is now more
generally reusable. Any existing code that was poking at Query.having
will now break in very visible ways (no subtle miscalculations, which is
a good thing).
This patch, en passant, removes the existing "having" test, since the
new implementation requires more setting up than previously. The
aggregates support (currently in a separate codebase) has tests for this
functionality that work as a replacement for the removed test.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9700 bcc190cf-cafb-0310-a4f2-bffc1f526a37
that happens with MySQL when a "GROUP BY" clause is included. This is a
backend-specific operation, so any other databases requiring similar
encouragement can have a function added to their own backend code.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9637 bcc190cf-cafb-0310-a4f2-bffc1f526a37
backend.
This allows Querysets to be cached for Oracle and should provide a model for
adding pickling support to other (external) database backends that need a
custom Query class.
Thanks to Justin Bronn for some assistance with this patch.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9272 bcc190cf-cafb-0310-a4f2-bffc1f526a37
again later (order_by('foo')). Or, at least, it can now. Thanks to Ilya
Novoselov for diagnosing the problem here.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9206 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Any extra(select=...) columns can be ignored in the SQL for dates, since we are
only interested in extracting distinct date values. We were previously
including them by accident and it was generating incorrect SQL.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9091 bcc190cf-cafb-0310-a4f2-bffc1f526a37
"having" attributes, only the former was included in the resulting SQL, meaning
subclasses had to completely duplicate Query.as_sql() if they were using any
kind of grouping filtering on the results.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@9007 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This was triggered by r8794, but was, in fact, fairly fragile before then. The
current fix is the correct way we should be doing this.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8898 bcc190cf-cafb-0310-a4f2-bffc1f526a37
sometimes also sharing aliases, instead of creating their own. This was
generating incorrect SQL.
No representative test for this fix yet because I haven't had time to write one
that fits in nicely with the test suite. But it works for the monstrous example
in #8790 and a bunch of other complex examples I've created locally. Will write
a test later.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8853 bcc190cf-cafb-0310-a4f2-bffc1f526a37
efficient than possible SQL in some odd cases (found via code inspection, not
any particular failing example).
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8831 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Yes, this really is a commit that fixes an oversight in a commit that fixed an
oversight. One day I'll get it right.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8829 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Also comes with approximately 67% less stupidity in the table joins for
filtering on generic relations.
Fixed#5937, hopefully for good, this time.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8644 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This isn't a complete solution to this class of problem, but it will do for
1.0, which only has generic relations as a multicolumn type. A more general
multicolumn solution will be available after that release.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8608 bcc190cf-cafb-0310-a4f2-bffc1f526a37
fields no longer creates duplicate copies of the join table(s). Basically, this
means filters on the join table (for ManyToManyField(through=...)) and complex
filters in the normal (non-through) case don't produce incorrect or duplicate
results.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8472 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Though some attempts and backwards-compatibility were made, speed trumped compatibility. Thus, as usual, check BackwardsIncompatibleChanges for the complete list of backwards-incompatible changes.
Thanks to Jeremy Dunck and Keith Busell for the bulk of the work; some ideas from Brian Herring's previous work (refs #4561) were incorporated.
Documentation is, sigh, still forthcoming.
Fixes#6814 and #3951 (with the new dispatch_uid argument to connect).
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8223 bcc190cf-cafb-0310-a4f2-bffc1f526a37
the prequisites are correctly initialised prior to using them. Only affects
Oracle and other db backends requiring resolve_columns() (e.g. MS SQL?)
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8112 bcc190cf-cafb-0310-a4f2-bffc1f526a37
match nothing). This allows for some more straightforward code in the admin
interface.
Fixed#7488 (all the debugging there was done by Brian Rosner, who narrowed it
down to the item in this patch).
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8061 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Based on a patch from Justin Bronn.
The test in this patch most likely breaks on Oracle. That's another issue.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@8053 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This removes some of the leaky abstraction problems (lifting WhereNode
internals into the Query class) from that commit and makes it possible for
extensions to WhereNode to have access to the field instances. It's also
backwards-compatible with pre-[7773] code, which is also better.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7835 bcc190cf-cafb-0310-a4f2-bffc1f526a37
This avoids any use of "pk is not NULL" fragment, which behave inconsistently
in MySQL. Thanks to Russell Keith-Magee for diagnosing the problem and
suggesting the easy fix.
Refs #7076.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7812 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Also added a section to the documentation to indicate why it's probably not a
good idea to rely on this feature for complex stuff. Garbage in, garbage out
applies even to Django code.
Thanks to erik for the test case for this one.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7791 bcc190cf-cafb-0310-a4f2-bffc1f526a37
We no longer store any reference to Django field instances or models in the
Where node. This should improve cloning speed, fix some pickling difficulties,
reduce memory usage and remove some infinite loop possibilities in odd cases.
Slightly backwards incompatible if you're writing custom filters. See the
BackwardsIncompatibleChanges wiki page for details.
Fixed#7128, #7204, #7506.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7773 bcc190cf-cafb-0310-a4f2-bffc1f526a37
so that the ordering doesn't accidentally restrict the result set.
(Ironically, one existing test actually showed this problem, but I was too
dumb to notice the result was incorrect.)
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7761 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Avoids joining with the wrong tables when connecting select_related() tables to
the main query. This also leads to slightly more efficient (meaning less tables
are joined) SQL queries in some other cases, too. Some unnecessary tables are
now trimmed that were not previously.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7741 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Previously, if we were querying across a nullable join and then a non-nullable
one, the second join would not be a LEFT OUTER join, which would exclude
certain valid results from the result set.
This is the same problem as [7597] but for values() field specifications, so
this covers the second case where Django adds extra stuff to the select-clause.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7740 bcc190cf-cafb-0310-a4f2-bffc1f526a37
database level. Also worked around the fact that MySQL (and maybe other
backends we don't know about) cannot select from the table they're updating.
Fixed#7095.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7496 bcc190cf-cafb-0310-a4f2-bffc1f526a37
marked as erroneous. It's just more dangerous and risky, not forbidden. This
commit restores backwards compatibility there.
git-svn-id: http://code.djangoproject.com/svn/django/trunk@7490 bcc190cf-cafb-0310-a4f2-bffc1f526a37
Anssi Kääriäinen
Tai Lee
Luke Plant
Michael Manfre
Malcolm Tredinnick
Alex Gaynor
Mitar
Aymeric Augustin
Alex Gaynor
Nuno Maltez
Claude Paroz
Adrian Holovaty
Julien Phalip
Jannis Leidel
Carl Meyer
Ramiro Morales
Karen Tracey
Russell Keith-Magee
Jacob Kaplan-Moss
Ian Kelly
Justin Bronn
Gary Wilson Jr