Justin, I just found a case when we have to make your LinkedHashMap fix (or other change to respect iteration order). In unit tests, if a complex type is programmatically constructed, the XSD schema that is automatically generated from it uses the property iteration order. This is effectively random with the current HashMap implementation, making it hard to predict what XML is encoded when a new property is added.
As an alternative to LinkedHashMap, perhaps we should hang onto the construction-time properties list so we can do iteration, equals, and hashCode in one consistent change? Maybe pulling up the implementation I put in SimpleFeatureTypeImpl? Opinions? Regards, Ben. Ben Caradoc-Davies wrote: > Justin, I have no objection. It should do no harm. However, note that: > > (1) order should never be important, except for special case subclasses > such as SimpleFeatureTypeImpl (see below), and > > (2) LinkedHashMap conforms to the Map contract, which guarantees that > order is *not* significant for equals/hashCode. Fixing iteration order > does not fix this. > > I experimented with using LinkedHashMap in ComplexTypeImpl to fix > SimpleFeatureTypeImpl iteration order, but when I realised it did not > fix equals/hashCode, I made all my changes to SimpleFeatureTypeImpl. If > you support my position, please join me in nagging Jody to get my patch > accepted. Please take a look at it; the patch comes with unit tests for > iteration order consistency: > http://jira.codehaus.org/browse/GEOT-2338 > > More importantly, why do you want to change the order of properties? Do > you have a non-simple use case in which they matter? If not, please > consider my patch. > > Kind regards, > Ben. > > > Justin Deoliveira wrote: >> Sorry, i did not mean tree map, just a map with predictable iteration >> order. LinkedHashMap should do the trick. >> >> Justin Deoliveira wrote: >>> Hi all, >>> >>> I think Ben may have brought this up recently, but looking at >>> ComplexTypeImpl it seems property values are stored in a hash map, which >>> totally throws away the order in which property descriptors are added to it. >>> >>> Any objection to making this a tree map as to maintain order? > > -- Ben Caradoc-Davies <ben.caradoc-dav...@csiro.au> Software Engineer, CSIRO Exploration and Mining Australian Resources Research Centre 26 Dick Perry Ave, Kensington WA 6151, Australia ------------------------------------------------------------------------------ Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H _______________________________________________ Geotools-devel mailing list Geotools-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/geotools-devel
