KarensProgress - JSOC

Karen's archive

. 02-28-2008

Phil helped me to gain more understanding of his naming document. In a dataset name, if none but ?? query exists, then we don't filter the latest version. I am looking into implementing this. I don't feel very comfortable with dataset names that involves multiple series, separated by commas. I am afraid some places are going to break. I am planning some tests.
Did rather laborious work to change owner/permission for namespace mdi from rick to dsdsdata. In particular, ownership change involves ALTER TABLE which requires an ExclusiveLock on the table. All of our running modules use AccessShareLocks on the master tables drms_*, hence they can block such ownership change. Another reason not to have long running transactions.
Worked with Art and Jim to fix a memory problem in sum_open.c. The problem has to do with freeing the memory that stores wd. sum_close() is not the place to free this memory. This is because between each pair of sum_open() and sum_close(), there will be a number of sum_get(), each of them allocates new memory. The proper place to free this memory is after sum_get() by any program that calls it, be it DRMS or not.
Merged in code for new "DBIndex" in jsd. "DBIndex" follows exactly the same syntax as "Index" (for specifying prime key).The max number of prime keys is set to 15 and max number of db indexed keyword is 10. For backward compatibility, when "DBIndex" is missing in a jsd, if the number of prime keys is no more than 5, then we create a composite index for all prime keys. This is the only case we'd create a composite index. Note it's legal for "DBIndex" to stand by itself without any keyword, no index will be created in that case.
Writing documentation

. 02-21-08