Re: [Cdk-devel] synchronized Vectors

[Egon W. <ewi...@un...>]

On Thursday 30 November 2006 13:47, Thorsten Meinl wrote:
> When trying to integrate an SDF parser into our data analysis platform
> KNIME, I had problems parsing molecules in parallel. The reason in this
> specific case is, that the IsotopeFactory uses a Vector for data that is
> read-only. Because a Vector is synchronized the different threads all
> block on the Vector which severly slows down the parsing. In such cases
> it's much better to use an ArrayList because it is not synchronized.
> There may be other classes that cause the same problem. A proposed patch
> for the IsotopeFactory is attached.

I do not think it really needs synchronization, and I think this is still one 
of those historic things in the CDK; I believe this code predates the 
introduction of ArrayList.

Anyway, I will apply this patch, and commit if it does not introduce 
regressions. Actually, I might very well look at that reader.readIsotopes() 
method too, and have that return a List instead of a Vector too.

Egon

-- 
CUBIC
blog: http://chem-bla-ics.blogspot.com/