That was fixed in June with UIMA-2418 (which I see I forgot to close) but is not in a release yet. I used LinkedHashMaps to preserve the order so that users also have some control over initialization order. I added JUnit tests that also verify Marshall's updates to preserve white-space and comments,
~Burn On Wed, Sep 19, 2012 at 2:43 AM, Ingo Thon <[email protected]> wrote: > Dear List, > > I’m using the Eclipse Plugin to edit UIMA-Descriptor Files. The resulting > files are stored and tracked in a mercurial repository. > However, when the files are opened and saved on different machines the > order of the entries in the resulting xml files are quite often slightly > reshuffled, e.g., > the order of the delegateAnalysisEngine. Same problem holds for the > different parameter. > Even though this doesn’t change the meaning of the descriptor it results > quite often in very annoying merge conflicts. > > Is there any way to change this behavior? > > As the plugin always reshuffles the same blocks of entries, it seems to me > that the XML File is stored in a HashMap and the hash computation > differs between those machines. If this is the case it should be possible > to use a SortedHashMap or something similar > > > Best regards, > Ingo
