Pyreases doesn't like tables names public*

[reedstrm]

Pyrseas$ git describe --tags
v0.7.1-47-g800158e

Pyrseas$ dbtoyaml test
Traceback (most recent call last):
  File "/opt/dev/bin/dbtoyaml", line 9, in <module>
    load_entry_point('Pyrseas==0.7.1', 'console_scripts', 'dbtoyaml')()
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/dbtoyaml.py", line 49, in main
    dbmap = db.to_map()
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/database.py", line 268, in to_map
    self.from_catalog()
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/database.py", line 178, in from_catalog
    self._link_refs(self.db)
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/database.py", line 138, in _link_refs
    db.triggers)
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/dbobject/table.py", line 822, in link_refs
    table.owner_column = self[(sch, table.owner_table)]. \
KeyError: (u'public', 'tions')

Pyrseas$ pg_dump -s test | grep -A 4 CREATE\ TABLE
CREATE TABLE publications (
    id integer NOT NULL,
    doc text
);

Something's eating my publications table's name

[jmafc]

That table by itself doesn't appear to cause a problem, at least when I test against a previously empty PG 9.1 database.

What I find weird is that the last line in the traceback is reported as being line 822, but I see that line at https://github.com/perseas/Pyrseas/blob/master/pyrseas/dbobject/table.py#L819. Did you perhaps edit the file?

[reedstrm]

This is a 9.3 empty db. That is weird - the one I see in front me me after doing a reinstall says 819:

git diff
(dev)reedstrm@desert:/opt/dev/Pyrseas$ git status
On branch master
Your branch is up-to-date with 'origin/master'.

nothing to commit, working directory clean
$ git remote -v
origin  https://github.com/perseas/Pyrseas.git (fetch)
origin  https://github.com/perseas/Pyrseas.git (push)

Pyrseas$ dbtoyaml test
Traceback (most recent call last):
  File "/opt/dev/bin/dbtoyaml", line 9, in <module>
    load_entry_point('Pyrseas==0.7.1', 'console_scripts', 'dbtoyaml')()
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/dbtoyaml.py", line 49, in main
    dbmap = db.to_map()
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/database.py", line 268, in to_map
    self.from_catalog()
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/database.py", line 178, in from_catalog
    self._link_refs(self.db)
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/database.py", line 138, in _link_refs
    db.triggers)
  File "/opt/dev/local/lib/python2.7/site-packages/Pyrseas-0.7.1-py2.7.egg/pyrseas/dbobject/table.py", line 819, in link_refs
    table.owner_column = self[(sch, table.owner_table)]. \
KeyError: (u'public', 'tions')

[reedstrm]

ok. testing w/ another new db, but using template0 does not fail, so I've got something installed in template1 causing this.

[reedstrm]

Ah 'cause my example lied, I cut it from a pg_dump. The critical bit is having a serial column:

create table publications (id serial, doc text);

even against template0 is good enough to trigger it.

[jmafc]

OK, reproduced. Will check what's wrong later.

[reedstrm]

thanks. Makes sense, that bit of code is linking up sequences owned by the table.

[reedstrm]

Last bit of debug - it's specific to tables that match public.public* with a serial column. A private.private* works, as does other.public*

[reedstrm]

Here's the code at fault: https://github.com/perseas/Pyrseas/blob/master/pyrseas/dbobject/table.py#L70-L75