David,I do not see any other way than programmatically separating the "lines" of the SQL query. Is this doable ? Like, if it's a backup, it probably has a chance that each query is a line, or ?
If yes, then it should be pretty easy to use a LineNumberReader and feed each line as an SQL query... or do I mistake ?
Alternatively, we could tweak the sql taglib to actually read the SQL script and not load it as a string but this would mean to decompose the lines in an appropriate way. I do not know how that could be done. If you know of a generic way, let's open a jira issue and work on that, it'd be easy.
thanks paul Le 10 juil. 07 à 01:04, Karr, David a écrit :
I have a large Jelly-SQL script (about 3.8 mb) that just does SQL callsto insert rows into a database. Each row it inserts is pretty small, but it inserts a lot of rows (relatively). It currently inserts about 18000 rows. What I'm finding is that the script won't even run on Windows (2 gig process limit). It takes too much memory. A previousversion of the script only inserted about 11000 rows, and it was able to run on Windows. The vast majority of the script is generated by a Javaapp that processes a spreadsheet. While the script is running, I watch it in Task Manager, and I see the memory very slowly increasing. It runs for quite a while. It finally runs out of memory in the JVM and fails. Is there some strategy for building a script like this so it is more memory-efficient? The script currently has a top-level "j:jelly" element, imports a utility package (some dbutils), and then imports the generated portion of the script. I can temporarily work around this by building multiple top-level scripts that call separate pieces of the big script, but that'sannoying. I'm also trying to get this set up on a Unix system, to get alarger process size. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
smime.p7s
Description: S/MIME cryptographic signature