-
I am attempting to produce HISPID5 from a database that should match the schema closely. The identifier role "d" in my database, according to the SME, maps to "det." . The schema at trunk/HISPID5Extension.xsd does not have a "det." as a value in the enumeration, but it does have "del." . Is this a typo or a transcription error? All the other values map across fine.
2009-10-30 00:54:10 UTC in Biodiversity Information Group
-
I have not run the full regression test suite against these changes. There is a change in functionality - the iterator on ObjectToInteger no longer returns a Long. However, this is explicitly declared in the new Generic declaration, and the codebase would not compile if there was code rlying on this behaviour.
Still - the test should be run - I'll just have to look through the docs and find...
2009-08-14 06:54:51 UTC in eXist
-
Patch to add generic declarations to org.exist.util.hashtable. Changed iterator type to an enum, and the condition test to a switch. This permits the compiler to compile down to a table jump instruction ... although it may already have been doing so, I suppose.
2009-08-14 04:39:04 UTC in eXist
-
test cases for the various org.exist.util.hashtable classes attached. These test cases are needed before these classes get generificated.
One failure: Object2IntHashMap holds its values as Longs, rather than Integers, and this means that the values returned in the iterator are Longs. I don't know if any other code relies on this.
I also note that the hashtable code is using new Integer(...
2009-08-12 06:25:10 UTC in eXist
-
Seeing as we are going all CS101, I have written an implementation of SmoothSort. SmoothSort is supposed to have performance advantages when dealing with data that has chunks of sorted regions in it. Of course, this code is not actually used anywhere yet, but it may be that there are parts of the app that might benefit from having it.
Attached is source and diff.
2009-08-07 08:00:31 UTC in eXist
-
Ok. I have implemented (most of) Floy'ds algorithm and attached a new source file. I ahve also added a test method to the test class in a separate bug report. Lucky I did! While developing this, that new method discovered that the code I was writing didn't work when sorting an array of two elements :-O . But, I've attached the code, it passes the tests, but I haven't tried out any performance...
2009-08-06 07:41:45 UTC in eXist
-
1 - test suite organises tests using sub-suites
2 - added test for different sizes of test array, subregion, including the array-length 2 case.
3 - removed System.out.println()s.
2009-08-06 07:14:16 UTC in eXist
-
Ah - ok. Floyd's first improvement is to *not use* bubbleup when doing the initial makeheap. Well, we are already doing this, so that's cool.
The other thing is that "drop" parametr, which I don't understand yet ... ok, I get it. In the old sort, each "siftdown" stage takes two comparisons: you compare the node with the two child nodes. In Floyd's one, you move the vacant node all the way...
2009-08-06 04:06:58 UTC in eXist
-
Ok. I have had a look at your code and a look at the various wikipedia pages mentioned.
The pages mention that Floyd's improvements concern the filterup procedure, which we are not using. Perhaps the sort could indeed be improved by using filterup in that first pass rather than using the current method. I'll have a closer look.
But the problem in the HSort class is that heapsort on a...
2009-08-06 02:59:01 UTC in eXist
-
That's awesome. This is the first time I have contributed to a collaborative open-source project, and I'm very pleased that my test cases are helping someone else's work.
I was thinking of putting the generic declarations against the sort algorithms, but I decided that making one fix at a time was the best option, and I wasn't sure that it wouldn't break the build.
Shall I go ahead and attempt...
2009-08-06 00:46:17 UTC in eXist
