Greetings, all,
It is with a sigh of relief that I announce the completed draft of the Hackystat
Developer Guide chapter on sensor design:
<http://hackydev.ics.hawaii.edu/hackyDevSite/external/docbook/ch17.html>
Please take a look and let me know of any comments or suggestions you have for
improvements.
Checking the subversion log in Jira, I report to my horror that this chapter was first
worked on almost a year ago, in April 2006. There was then a break of 8 months until work
started again in January 2007. In the interim, the scheduled release for this issue was
deferred from 7.4 to 7.5 to 7.6 to 7.7.
The good news is that we're closing in on a relatively complete documentation set for the
basic facilities in Hackystat. We're at 28 chapters right now, and there's really only a
couple more unwritten chapters left (the Hackystat Test Framework, and Designing
Commands).
I am going to now take a welcome break from DocBook. I started work yesterday on
removing the hackySdt_WorkspaceMap module from the system. The workspace map mechanism
is another one of those design experiments that seemed like a good idea at the time, but
turns out to be problematic for a number of reasons:
- it's brittle and introduces sensor data collection dependencies. For example, UnitTest
data can't be analyzed unless FileMetric data is sent for the same day. That's very
unintuitive.
- it's expensive space-wise. The WorkspaceMap can consume a lot of heap and not even be
used.
- it's language-specific. Our current WorkspaceMap implementation works only
for Java.
Our experience with WorkspaceMap taught us that the better design is to simply require
all sensor data to include some kind of path information. We already updated the SDTs
over a year ago but kept the WorkspaceMap around for backwards compatibility. I think at
this point we are better off simply eliminating the mechanism and if someone wants to use
the old data, they will have to write an updater program that parses the XML of the old
data and rewrites it with the path attribute added.
It turns out that it was easy to remove the WorkspaceMap module, but the result is that a
bunch of test cases are now breaking because they rely on "old" sensor data without path
information. So, I'm going to need a few days to hack around in UnitTest, Coverage, and
Dependency SDT modules to bring the test data sets into the present.
Cheers,
Philip
