Posted: January 13th, 2012 - So I have been working on a script in OpenBD that loops over a large 1.2GB xml file and parses specific nodes. Since the file is so big I have to loop over the file line by line and build each node individually before running xmlparse() on the node. Everything was working just dandy until I got to around 6 million lines.

Looking at the task manager on the server I could see the memory gradually rising and eventually at around 7 million the script died. I had no clue what to do and tried several very unsuccesffuly modifications until an awesome person by the name of Jamie MacDonald on the OpenBD google group told me to try putting <cfflush /> at the end of my loop. The explanation given to me was that even though I was not outputting any data the whitespace characters inside the loop may be adding up because the CFML engine thinks it is needed to render a page and thus holds the whitespace in the buffer.

HUH!? Whitespace problems inside a component with output="false"? CFdocs.org says that output= false should make the code act like it is inside one big cfsilent, but maybe I am misunderstanding something. It could also be a bug specific to OpenBD, but maybe not. Anyway using cfflush allowed me to process the whole file with no memory problems. I watched the task manager while the script was running and with cfflush the memory stayed consitant at a normal level.  So to sum up, if you run into memory issues while using a long running cfloop try putting <cfflush /> at the very end of the loop. Hopefully this will be useful to people using the other CFML engines besides OpenBD too.