Out of _build_joins_for_select() itables_to_merge variable has only one
table which is the one implicated in the join (in this case "auth_user")...
But another merge_tablemaps() occurs in red below :
if join:
(
# FIXME? ijoin_tables is never used
ijoin_tables, ijoin_on, itables_to_merge, ijoin_on_tables,
iimportant_tablenames, iexcluded, itablemap
) = self._build_joins_for_select(tablemap, join)
tablemap = merge_tablemaps(tablemap, itables_to_merge)
tablemap = merge_tablemaps(tablemap, itablemap)
if left:
(
join_tables, join_on, tables_to_merge, join_on_tables,
important_tablenames, excluded, jtablemap
) = self._build_joins_for_select(tablemap, left)
tablemap = merge_tablemaps(tablemap, tables_to_merge)
tablemap = merge_tablemaps(tablemap, jtablemap)
So itables_to_merge get altered and get back to is previous state as it is
at some point in _build_joins_for_select() (in red), but it was filtered in
the stage in blue :
def _build_joins_for_select(self, tablenames, param):
if not isinstance(param, (tuple, list)):
param = [param]
tablemap = {}
for item in param:
if isinstance(item, Expression):
item = item.first
key = item._tablename
if tablemap.get(key, item) is not item:
raise ValueError('Name conflict in table list: %s' % key)
tablemap[key] = item
join_tables = [
t._tablename for t in param if not isinstance(t, Expression)
]
join_on = [t for t in param if isinstance(t, Expression)]
tables_to_merge = {}
for t in join_on:
tables_to_merge = merge_tablemaps(tables_to_merge,
self.tables(t))
join_on_tables = [t.first._tablename for t in join_on]
for t in join_on_tables:
print t
if t in tables_to_merge:
print t
tables_to_merge.pop(t)
important_tablenames = join_tables + join_on_tables + \
list(tables_to_merge)
excluded = [
t for t in tablenames if t not in important_tablenames
]
return (
join_tables, join_on, tables_to_merge, join_on_tables,
important_tablenames, excluded, tablemap
)
With the logic in merge_tablemaps() where things are reversed in big and
small and big get update at the end it explains why itables_to_merge get
modified by merge_tablemaps()
def merge_tablemaps(*maplist):
"""Merge arguments into a single dict, check for name collisions.
Arguments may be modified in the process."""
ret = maplist[0]
for item in maplist[1:]:
if len(ret) > len(item):
big, small = ret, item
else:
big, small = item, ret
# Check for name collisions
for key, val in small.items():
if big.get(key, val) is not val:
raise ValueError('Name conflict in table list: %s' % key)
# Merge
big.update(small)
ret = big
return ret
I don't think itables_to_merge is updated intentionally because why
filtering it in _build_joins_for_select()...
My guess is the intention wasn't to get it updated and the only purpose of
the tablemap with all the tables names as itablemap use lack some(s) of
them...
So deepcopy would be need it or just dict(**itables_to_merge)) while
passing itables_to_merge to merge_tablemaps()
*Should we fix that??*
if join:
(
# FIXME? ijoin_tables is never used
ijoin_tables, ijoin_on, itables_to_merge, ijoin_on_tables,
iimportant_tablenames, iexcluded, itablemap
) = self._build_joins_for_select(tablemap, join)
tablemap = merge_tablemaps(tablemap, dict(**itables_to_merge))
tablemap = merge_tablemaps(tablemap, itablemap)
if left:
(
join_tables, join_on, tables_to_merge, join_on_tables,
important_tablenames, excluded, jtablemap
) = self._build_joins_for_select(tablemap, left)
tablemap = merge_tablemaps(tablemap, dict(**tables_to_merge))
tablemap = merge_tablemaps(tablemap, jtablemap)
On Tue, Feb 27, 2018 at 4:30 PM, Richard Vézina <[email protected]
> wrote:
> In 17.01
>
> line 627
>
> if join and not left:
> cross_joins = iexcluded + list(itables_to_merge)
>
> cross_joins contains both tables involve in the query, which wasn't the
> case in 16.11
>
> I follow up and it could come from
>
> helpers.methods.merge_tablempas() (introduced in 17.01) use
> in _build_joins_for_select() or it could be caused by this piece of code
> included in the former mention function :
>
> tablemap = {}
> for item in param:
> if isinstance(item, Expression):
> item = item.first
> key = item._tablename
> if tablemap.get(key, item) is not item:
> raise ValueError('Name conflict in table list: %s' % key)
> tablemap[key] = item
>
> At that point code get difficult to understand what it does...
>
> Should we consider this a flaw in the actual code or is this issue which
> goes away when not doing empty query is related to the difficulty we were
> having supporting the empty query syntax???
>
> I guess the answer to this question will resolve this definitely...
>
> Richard
>
> On Tue, Feb 27, 2018 at 2:11 PM, Richard Vézina <
> [email protected]> wrote:
>
>> What causing the CROSS JOIN issue had been introduce between pyDAL 16.11
>> and 17.01
>>
>> There is a compare of the 2 : https://github.com/web2py/py
>> dal/compare/v16.11...v17.01
>>
>> Richard
>>
>> On Tue, Feb 27, 2018 at 12:59 PM, Richard Vézina <
>> [email protected]> wrote:
>>
>>> At least there is a open issue on the book repo :
>>> https://github.com/web2py/web2py-book/issues/346
>>>
>>> On Tue, Feb 27, 2018 at 12:57 PM, Richard Vézina <
>>> [email protected]> wrote:
>>>
>>>> Here : https://github.com/web2py/pydal/issues/388
>>>>
>>>> I think the follow up update to the book and change log never occurs...
>>>>
>>>> I am fine with patching my code and make sure I don't leave empty
>>>> query... But I am curious to know if this "CROSS JOIN table_name" addition
>>>> could be avoid otherwise than providing non empty query... I mean it rough
>>>> mistake...
>>>>
>>>> Richard
>>>>
>>>> On Tue, Feb 27, 2018 at 12:39 PM, Richard Vézina <
>>>> [email protected]> wrote:
>>>>
>>>>> @Leonel, I recall about empty query beeing deprecated... But I can't
>>>>> find any reference to it neither in change log of pyDAL nor web2py change
>>>>> log... Niether the book...
>>>>>
>>>>> Which version of the DAL the support been dropped?
>>>>>
>>>>> Richard
>>>>>
>>>>>
>>>>>
>>>>> On Tue, Feb 27, 2018 at 12:01 PM, Richard Vézina <
>>>>> [email protected]> wrote:
>>>>>
>>>>>> In [6]: db(db.auth_user.id > 0)._select(db.auth_membership.group_id,
>>>>>> join=db.auth_membership.on(db.auth_user.id ==
>>>>>> db.auth_membership.user_id))
>>>>>> Out[6]: 'SELECT "auth_membership"."group_id" FROM "auth_user" JOIN
>>>>>> "auth_membership" ON ("auth_user"."id" = "auth_membership"."user_id")
>>>>>> WHERE
>>>>>> ("auth_user"."id" > 0);'
>>>>>>
>>>>>> Seems to solve it...
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Tue, Feb 27, 2018 at 11:58 AM, Leonel Câmara <
>>>>>> [email protected]> wrote:
>>>>>>
>>>>>>> I believe we deprecated empty queries as they caused way too many
>>>>>>> problems and bugs (e.g. common filters don't work).
>>>>>>>
>>>>>>> --
>>>>>>> Resources:
>>>>>>> - http://web2py.com
>>>>>>> - http://web2py.com/book (Documentation)
>>>>>>> - http://github.com/web2py/web2py (Source code)
>>>>>>> - https://code.google.com/p/web2py/issues/list (Report Issues)
>>>>>>> ---
>>>>>>> You received this message because you are subscribed to the Google
>>>>>>> Groups "web2py-users" group.
>>>>>>> To unsubscribe from this group and stop receiving emails from it,
>>>>>>> send an email to [email protected].
>>>>>>> For more options, visit https://groups.google.com/d/optout.
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>
--
Resources:
- http://web2py.com
- http://web2py.com/book (Documentation)
- http://github.com/web2py/web2py (Source code)
- https://code.google.com/p/web2py/issues/list (Report Issues)
---
You received this message because you are subscribed to the Google Groups
"web2py-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
