cextension does not allow list-based rows from the pyodbc DBAPI

[Mike Bayer]

the C version of RowProxy has an assertion that the rows returned by fetchone() and similar must be a tuple. DBAPI specifies a "sequence" as the row type and only suggests a tuple. PyODBC returns rows as lists and thus the C extensions are currently unusable with PyODBC.

[Former user]

(original author: ged) I'll handle this...

[Former user]

(original author: ged) fixed in 7e25d8218bee860e043a73ed1c9d887ed9920855

[Former user]

Unfortunately, all my test cases use a Vertica instance which is a pain in the ass to acquire. I will try to generate some test cases using PyODBC and a more available database. In the meantime, I'll post these stack traces in the hopes that they may be helpful.

(Also, not sure if this should go in a new ticket or not. Please advise if I'm violating the Process here.)

====================================================================== ERROR: testTransactionIsolation (main.TestVerticaDialectConnection)

---

Traceback (most recent call last): File "reporting/test/test_vertica.py", line 57, in testTransactionIsolation _, iso_level = e.execute('SHOW TRANSACTION_ISOLATION').fetchone() RuntimeError: number of values in row (26057472) differ from number of column processors (2)

====================================================================== ERROR: testGetColumns (main.TestVerticaDialect)

---

Traceback (most recent call last): File "reporting/test/test_vertica.py", line 106, in testGetColumns self.engine, self.test_tbl, schema='public') File "<string>", line 1, in <lambda> File "/home/vmc/ENV/lib/python2.6/site-packages/SQLAlchemy-0.6beta3dev-py2.6-linux-x86_64.egg/sqlalchemy/engine/reflection.py", line 32, in cache return fn(self, con, args, *kw) File "build/bdist.linux-x86_64/egg/reporting/vertica/base.py", line 121, in get_columns pk_columns = [x0 for x in connection.execute(spk)] SystemError: ../Objects/tupleobject.c:101: bad argument to internal function

[Former user]

(original author: ged) Replying to guest:

Unfortunately, all my test cases use a Vertica instance which is a pain in the ass to acquire. I will try to generate some test cases using PyODBC and a more available database.

Please do.

Traceback (most recent call last): File "reporting/test/test_vertica.py", line 57, in testTransactionIsolation _, iso_level = e.execute('SHOW TRANSACTION_ISOLATION').fetchone() RuntimeError: number of values in row (26057472) differ from number of column processors (2)

Something seem to be awfully wrong with that dialect...

Could you try to replace in your test:

_, iso_level = e.execute('SHOW TRANSACTION_ISOLATION').fetchone()

by:

proxy = e.execute('SHOW TRANSACTION_ISOLATION').fetchone()
row = proxy._row
print type(row), len(row), row

and paste the result here?

[Former user]

(original author: ged) By the way, does your test case run correctly without the C extension?

[Former user]

(original author: ged) Replying to ged:

By the way, does your test case run correctly without the C extension?

Ok, I saw the thread on the list. So this ought to be a bug in the ext. I'll investigate this further...

[Former user]

(original author: ged) Ok, I figured it out, but the simple fix to that (s/Py_SIZE/PySequence_Length) slows things down a bit. I'll see if I can come up with a fix which doesn't. I'm too tired to do this today though.

[Former user]

Hi ged,

Thanks for your work on this. These tests do pass without the c ext. enabled.

The code you instructed me to run resulted in:

<type 'pyodbc.Row'> 2 ('transaction_isolation', 'READ COMMITTED')

The pyodbc Row class appears to be implemented as a pure C extension:

http://code.google.com/p/pyodbc/wiki/Rows

http://github.com/mkleehammer/pyodbc/blob/master/src/row.cpp

[Former user]

Oh, one other question which I forgot to ask - are the two errors above related to the same issue or is it two separate issues?

[Former user]

(original author: ged) Replying to guest:

Oh, one other question which I forgot to ask - are the two errors above related to the same issue or is it two separate issues?

RuntimeError?: number of values in row (26057472) differ from number of column processors (2)

This one is the same issue as the initial issue. I just forgot to apply the change in some places in my first patch. As for the second traceback, I don't know. It is probably the same issue, but I can't say for sure. I'll fix the first one and we'll see if that fixes the second one.

[Former user]

(original author: ged) Should be fixed in d3bc6c08d4484538171456cd4253bb3d47fa99e0.

[Former user]

(original author: ged) A test case for the changes

[Former user]

(original author: ged) Closing for now. Please reopen again if it still doesn't work.

[Former user]

Works great on my Vertica test cases now. Thanks.

[Mike Bayer]

Removing milestone: 0.6.0 (automated comment)