james-server-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefano Bagnara <apa...@bago.org>
Subject Re: [Mime4j] Benchmark of 0.3 vs 0.4
Date Wed, 06 Aug 2008 08:15:43 GMT
Niklas Therning ha scritto:
> Bernd Fondermann wrote:
>> On Mon, Aug 4, 2008 at 13:51, Niklas Therning <niklas@trillian.se> wrote:
>>  
>>> Hi,
>>>
>>> I've done some benchmarking of the recent performance improvements in 
>>> trunk
>>> vs Mime4j 0.3. I must say that the result is very promising! :)
>>>     
>>
>> <snip/>
>>
>>  
>>> Please let me know if my benchmarking method is flawed in any way. I'm
>>> attaching the source code of the test tool.
>>>     
>>
>> Did you do consistence checks by looking at what was actually returned
>> when parsing? For example, if the parser throws exceptions early, it
>> will indeed be faster than if it in fact processes all data. ;-)
>>
>>   
> 
> Yes I did that some weeks ago, after Oleg's performance patches. This 
> test dumped the structure of all 130000 messages to XML files and also 
> decoded all base64 and qouted printable parts into separate files. After 
> some bugs were fixed I got almost exactly the same results as with 0.3. 
> The only difference was that some messages (about 10 out of the 130000) 
> had an extra newline in the parsed epilogue or preamble. I don't think 
> that is an issue so I ignored those differences.

It is really good that you have a similar message base prepared to run 
with mime4j.

In current mime4j we added some more test to the testsuite that was 
checking the parsing. We now also check writing out the message and 
roundtrip read/write it.

Would you mind running the test (ExampleMessagesRoundtripTest) and see 
if you messages failing the roundtrip tests? I would be interested in 
them. You can even send them to me directly and I'll try to create new 
messages to reproduce the issues.

The first time you run this test it will create .mime4j.out files that 
you have to rename to .out (theoretically after checking they are really 
the expected output, but in this case you have 130000 messages and we 
can ignore this step simply caring of roundtrip for current output).
If you change "msgoutFileMime4j" variable just before the first run it 
will create .out files so you don't have to rename .mime4j.out to .out 
after the first run. Once you have out file you run again the test suite 
and see any error/failure.

About the newline I would be interested in evaluating this difference. 
If you find the time please send me one of this messages so that we can 
check if it was ok in 0.3 or it is ok in 0.4 (or both).

Thank you,
Stefano


---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org


Mime
View raw message