Page 3 of 5
The initial sketch did not bring over the my idea very well, so it was received with very little enthusiam. The version you see was made in Inkscape, which I can’t recommend often enough as excellent free (as in freedom) vector graphics tool. I just imported a scan of my initial sketch and handtraced it with splines and stuff..
I still like the result..
This blog has been relatively stale for a while now — mostly due to increasing work pressure and the limited amount of time I had left for it. What little time I had was mostly wasted on coding on it. The previous version was written in turbogears which seemd like a good idea back in 2006. Over time I kind of regretted that decision, which was not really Turbogears fault, but just caused by my laziness to really update it etc.
I’ve been frustrated with the old blog code and the old design for some time now and wanted to redo everything and already started multiple attempts in various technological reincarnations — the latest being an implementation in java/spring using my jcouchdb library. The design got more and more ambitious over time; so ambitious that it would basically require me to work full-time on it. Continue reading
Implementing a streaming attachment feature for jcouchdb, I started to wonder whether it would be a good idea for svenson to support JSON parsing from a stream, too, as I don’t really need the complete stream to start constructing the java object graph.
Implementing stream parsing was really nice and easy thanks to the units test present in svenson. After that, I came upon two ways to generally cut down on memory use. All tokens with fixed values could just have a single instance. The recording of tokens to provide token based look ahead was not really needed in all cases. But how much does that save?
As a test case I wrote a small tool class to generate random, nested JSON datasets, generated two test files of 65kb and 4.5mb size and parsed these with svenson 1.2.8 and what now is svenson 1.3.
Measuring the actual memory usage for these two test files proved to be difficult. Somehow none of the programs I tried seemed to give me the data I wanted. Eclipse TPTP just ignored Strings that were no member of any class but just parameters, making stream and string parsing look exactly the same memory-wise. tijmp and others did not provide the data I wanted at all.
So in the end I wrote a little python script that parses a hprof ASCII output to
- sum up all memory use
- group allocations by class type, but only if the stack trace of it touches svenson
- output the top 10 of those classes and the sums
This provided meaningful data and also showed some points for further improvement. There was a huge number of java.lang.reflect.Method allocations which turned out to be caused by svenson inspecting the target classes for annotations and appropriate methods which was done on a per target basis instead of the better per target class basis.
All in all the memory usage went down quite a bit:
45% less memory for the small file and 62% for the large file for all allocations. I think that is really good..
Below are some links to the files needed to repeat the benchmarking. The transform hprof script might also prove to be useful for other projects if changed appropriately.
The new jcouchdb release will also use stream parsing.
- Detailed data: Top 10 memory use by class and sums for all 6 variants
- Python script to transform hprof ASCII output
- Random JSON Generator
- Small JSON test file
- Big JSON test file
The command to generate the hprof file was something like
java -agentlib:hprof=heap=sites,depth=100,cutoff=0 -cp .. svensonperf.ReadJSONOld big.json
After again spending time to fix issues resulting from a collision of the yuicompressor jar and the normal rhino jar in one of my projects, I came up with a more radical solution:
- Download the source codes from yuicompressor and the corresponding rhino release
- Replace every occurance of “mozilla” with “yuizilla”
So now I have a version of the yuicompressor that works fine and does not conflict with the rhino version I also have in my project. And I don’t need any stupid jar class loaders or have to write stdin/stdout handling for some terribly slow external yuicompressor process. I can just use the classes
and be done with it. Hurray for fast dynamic server-side script and style compression!