Psycopg psycopg → psycopg2 2.5.3

Comments and changes to this ticket

• 

  Federico Di Gregorio September 20th, 2013 @ 02:05 PM

  If I remember correctly the libpq docs don't say if it is possible to
  retrieve the number of inserted columns and we're not sending data
  line-by-line but as chunks so counting the lines isn't possible. Maybe
  (a big maybe) the final PGresult contains that but I can't find this documented anywhere. We'll need to instrument psycopg to print PGresult
  fields and do a bit of guesswork.

• You flagged this item as spam.
  

  Rui Da Costa December 9th, 2013 @ 01:49 PM

  I'm no expert, but isn't it described here?:
  www.postgresql.org/docs/9.3/static/libpq-exec.html#LIBPQ-PQCMDTUPLES

  This function returns a string containing the number of rows affected by the SQL statement that generated the PGresult. This function can only be used following the execution of a SELECT, CREATE TABLE AS, INSERT, UPDATE, DELETE, MOVE, FETCH, or COPY statement, ...

• You flagged this item as spam.
  

  Nicholas Chammas May 5th, 2014 @ 09:59 PM

  On a related note, I have been using psycopg2 with Amazon Redshift and recently asked a similar question about getting the rowcount for COPY statements executed via cursor.execute. I've attached a PDF of the linked AWS forum page in case you can't access it.

  Amazon claims that Redshift correctly reports rowcount to the calling interface and that psycopg2 doesn't seem to pick it up.

  How can I validate Amazon's claim?

  • AWS_Developer_Forums__Does_Redshift_report_a_-1_rowcount_for_...pdf 96.1 KB
• 

  Daniele Varrazzo May 5th, 2014 @ 11:00 PM

  • State changed from “open” to “resolved”
  • Milestone set to “psycopg2 2.5.3”

  As Rui pointed out the number of rows is returned by PQcmdTuples. We weren't calling this function on copy results but only after normal opreations.

  Fixed in 2.5.3. Nicholas, try checking out https://github.com/psycopg/psycopg2/tree/maint_2_5 and see if it works as expected, thank you.

• You flagged this item as spam.
  

  Bilal Aslam May 27th, 2014 @ 06:02 PM

  Daniele, thanks for checking in the fix. I tried this out on May 27 using the branch you suggested, but I am still getting a rowcount of -1.

  In [36]: cursor.rowcount
  Out[36]: -1

  Nicholas, did you have any better luck than me?

• 

  Daniele Varrazzo May 27th, 2014 @ 07:32 PM

  Hello Bilal,

  yes, it works for me: check the test suite. The changeset is released in psycopg 2.5.3

  Note that old postgres versions (before 8.2) don't return the rowcount. So it depends on a message the server must send. If the server doesn't send it we cannot help.

• You flagged this item as spam.
  

  Nicholas Chammas May 27th, 2014 @ 10:09 PM

  Hey Daniele,

  It appears that we are still seeing the -1 rowcount on COPYs in psycopg 2.5.3 when testing against Redshift 1.0.789. This may well be an issue on Redshift's side, and I think it's beyond psycopg2's mandate to work cleanly with forks of PostgreSQL. (I think Bilal is also testing against Redshift.)

  FYI: Redshift is a distant fork of PostgreSQL 8 with a column-oriented storage engine, hosted as a service by Amazon. Its primary use case is for massive analytics workloads. Amazon recommends using PostgreSQL drivers to interact with the database.

  Anyway, I am following up with Amazon support here to get to the bottom of this. They claim it should work.