Re: runtime.parser.Parser maintains multiple VelocityCharStreams?

I've no idea.  I've never really dug into those much.  If Jarkko is
kicking around, he might know.

On Fri, Apr 12, 2013 at 10:41 AM, Christopher Schultz
<[email protected]> wrote:
> All,
>
> I've been doing some profiling looking for wasted memory in my webapp
> and I've discovered that org.apache.velocity.runtime.parser.Parser
> contains a few members that appear to be (essentially) duplicates.
>
> First, there is the "velocharstream" member which is a
> VelocityCharStream. Then, there is also a token_source member which is a
> ParserTokenManager which has a separate VelocityCharStream
> ("velcharstream" member). YourKit reports that these two are distinct
> objects, and they don't appear to share anything so I think they really
> are distinct.
>
> Each VelocityCharStream has a pair of 4096-byte buffers ("bufcolumn" and
> "bufline") plus a 8192-byte buffer in the "inputStream" member (a
> java.io.InputStreamReader), so each Parser is currently eating up 16k of
> memory in buffers. Yes, I know it's really nothing, but I'm wondering if
> buffers (or streams) are being allocated when they don't need to be
> allocated.
>
> Does anyone know why there might be two separate VelocityCharStreams at
> the Parser level? How about the 2 bufline/bufcolumn buffers in
> VelocityCharStream as well as the 8192-byte InputStreamReader buffer?
>
> The code has no inline documentation of Javadoc to suggest the purpose
> of these members.
>
> Thanks,
> -chris
>