I can think of a couple that rdbGUI in its current incarnation doesn't handle well or at all. Are there others?
If 2. is the problem, it could perhaps be addressed by making some enhancements to rdbGUI so that it would remember values for those columns – most of them – which tend to be the same for a collection of calibrations.
| Field name | Sticky? | Default/Initial value |
|---|---|---|
| instrument | Yes | (none)/"LAT" |
| flavor | Yes | "vanilla" |
| calib_type | No | (none)/"TKR_DeadChan" |
| data_ident | No | (none) |
| data format | Usually | (none)/"XML" |
| vstart | Usually | (none) |
| vend | Usually | "2037-12-31" |
| input_desc | No | null |
| notes | No | null |
| proc_level | Usually | "TEST" |
| completion | Usually | (none)/"OK" |
| locale | Yes | "orbit" |
| fmt_version | Usually; especially now while it's not used in a serious way | null |
| prod_start | Yes. Normally null upon INSERT | null |
| prod_end | Yes. Normally null upon INSERT | null |
| data_size | Wouldn't be if used | null |
In the production database, there should be no more than one calibration of a particular calib_type of production quality valid for any instant of time. When entering a new calibration, the natural thing to do is to set vend to its default value (Dec. 31, 2037), however the same operation should also set the value of vend for the previous production calibration of that type to the vstart of the new one, so that there is no overlap. Updates as well as inserts can give rise to this situation; that is, can make it necessary to modify some other row or rows in order for the database to continue to satisfy some consistency condition. It would be well to also have a means of checking (triggered on demand rather than by an update or insert) whether the database satisfies some set of conditions.
Last modified: