Pete Erickson <redlamb@xxxxxxxxxxx> writes: > I am looking for some help regarding an python OperationalError that I > recently received while executing a python script using sqlalchemy and > psycopg2. The python script parses an xml file stored on a networked > drive and enters the information into a pgsql database. Sometimes > these xml files reference a binary file which is also located on the > networked drive. These files are subsequently read in and stored in a > table along with the file's md5. The binary data is stored within a > bytea column. This script worked pretty well until recently when it > came across a binary file about 258MB in size. While reading the file > off the networked drive I received an OperationalError indicating that > it was unable to allocate memory for the output buffer. My initial > guess was that it ran out of memory, but according to the task manager > the machine had close to 2GB free when the error occurred. Out of memory is probably exactly right. The textual representation of arbitrary bytea data is normally several times the size of the raw bits (worst case is 5x bigger, typical case perhaps half that). In addition to that you have to consider that there are likely to be several copies of the string floating around in your process' memory space. If you're doing this in a 32bit environment it doesn't surprise me at all that 258MB of raw data would exhaust available memory. Going to a 64bit implementation would help some, but I'm not sure that that's an available option for you on Windows, and anyway it doesn't eliminate the problem completely. If you want to process really large binary files you're going to need to divide them into segments. regards, tom lane -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
