[Helma-user] Java heap space in dbh.query()

[Joshua Paine]

Julian Tree wrote:
> "Java heap space" error.  

Look in Helma's start.bat or start.sh file. You'll find a line which may 
be commented out that says something like:

set JAVA_OPTIONS=-server -Xmx128m

Uncomment (if necessary) and try doubling the -Xmx value, e.g., to 
-Xmx256m. (And then try doubling again if necessary and if you have the 
RAM.) But if that does the trick it means you were running out of 
memory, and if that happens now in development it's probably likely to 
happen again with heavier traffic in production, even with an increased 
memory limit.

The two obvious approaches are:

1) Don't get 3000 records. Humans can't usually make use of so many 
anyway--is there a way you could use limit clauses to offer paged 
results or something instead?

2) In most environments, the database driver can return a result set 
which lets you work with one result row at a time without loading them 
all into memory. Helma's Database object is extremely rudimentary, 
however, and doesn't offer this feature (nor parameterized/prepared 
statements--it's really not meant for serious DB work since they're 
expecting you to use the built-in ORM for most things). You can get that 
feature by dropping down to Java or by directly managing Java's database 
access stuff through Rhino's Java/JavaScript interfaces (don't ask me 
how, I haven't the slightest) or you can simulate it by running the 
query with limit clauses in a loop to retrieve, e.g., 100 rows at a 
time, do your work with them, delete the array, and then get another 100.

> cachesize = 300000

This affects only the in-memory cache of HopObjects. If anything, 
increasing it will lead to *more* memory problems.

-Joshua