Storage Management Log
Gw
Grade: 6
This is nice log of the project, but lacks any reflection about the software engineering process, which was the goal.
-- EelcoVisser - 03 Jan 2005
Success! After a few long coding sessions, the Java side of the RA layer code is beginning to take shape. Today I even got the wiki in a state where it could open pages properly through RA code. I have not committed the code yet, since I still have a few issues to sort out. -- OlafJansen - 02 Nov 2004 We've decided not to use the RA code in the production JAR that will be produced later today, and which will be used in a code review and presentation of the project. Olaf and Patrick will work on the RA layer today, hopefully it will become feature complete by the end of this week (and the end of this project). Sander and I will start writing a project report, which will be used during the code reviews tomorrow. -- MichielRook - 01 Nov 2004 As Sander points out in his comment below here, we're in maximum crunch mode for the RA layer code. I have just added some directory-traversal niceties, and it looks like the RA code is coming along quite nicely. The big question is of course: will it all be done in time for acceptance testing? I think the big lines will be there, but there might still be some bugs in need of squashing. -- OlafJansen - 26 Oct 2004 There were some nix problems yesterday, so that was a good time to create missing JavaDocs. Today were on full steam again with the RA Layer. A lot of pointers out of the C-code have to be stored in the SVNJNIAdapter. First we opted for setters/getters for each pointer, but this is not very scalable. Reflection will be used for every RA-action, not very nice... So now, all the bookkeeping is done in C. Pointers are stored in a struct, and only the pointer to this struct is stored in Java code. User-management team is working in a seperate branch on our code, to add access-checking. -- SanderMak - 26 Oct 2004 At this point, functionality for getting files and commiting files to te RA layer is implemented. I am quite confident that we have a working RA layer storage next week. -- PatrickCloosterman - 24 Oct 2004 Today I have begun to dig deeply into the bowels of the SVN C API, and besides almost getting dizzy of the batons flying all over the place, I'm pretty much starting to get the hang of it. It was only after I implemented it myself that I took a look at Eelco's work, which turned out to be quite similar. D'oh. Well, at least I grok it now, and work on an RA-layer wrapper has started. I am quite confident that this approach is going to work (given enough coding time), but we do need to take a deep breath and explore the commit-functionality, which is rather complicated with a lot of complex delta-stuff going on. The bookkeeping Sander mentions below this post should also not be neglected, and could turn out to be a (surprisingly) large amount of work. -- OlafJansen - 18 Oct 2004 After some research our team opted out of SWIG bindings and the likes. Eelco provided us with a proof-of-concept for implementing RA functionality directly with the SVN C-api. This will be the way to go for now. The Storage functionality will be implemented in C directly. Then, with JNI this implementation will be called from Java. The advantages of this approach:
- No more checkouts needed
- (Somewhat buggy) Java subversion client API can be left behind
- a lot of bookkeeping will be our responsibility, now the WC-layer is abandoned
- this approach might need extensive caching