From: Daniele Varrazzo <daniele.varrazzo@gmail.com>
To: Ian <youknowho2000@yahoo.com>
Cc: "psycopg@postgresql.org" <psycopg@postgresql.org>
Sent: Friday, August 19, 2011 3:40 AM
Subject: Re: [psycopg] Reading binary data from Postgres 9
On Fri, Aug 19, 2011 at 1:28 AM, Ian <
youknowho2000@yahoo.com> wrote:
> Hi list,
> I am storing a NumPy array in binary form in Postgres with:
> import numpy as np
> ...
> my_array = np.array(data_list, np.float32)
> cur.execute("insert into my_table values (%s)",
> (psycopg2.Binary(my_array),))
Are you sure? I get an error with this (can't escape numpy.ndarray to
binary). Looks like it would be required to use
psycopg2.Binary(my_array.tostring()). Is it a typo from you or a
regression? (I'm using 2.4.x).
> I've confirmed that this works fine. However when I query this binary data
> from the database I am having trouble deserializing/getting it back into a
> NumPy array. The result comes back as a read-only buffer. Does psycopg2 have
> an opposite of Binary to handle this?
you can use numpy.frombuffer(b, np.float32) to have your data back,
where b is the buffer read from the cursor.
Cheers,
-- Daniele
