search

bradbell / at_cascade

Cascading Dismod_at Analysis From Parent To Child Regions
https://at-cascade.readthedocs.io
4 stars 3 forks source link

New wishlist items #20

Open ntemiq opened 1 month ago

ntemiq commented 1 month ago

Added 2 wishlist items

  1. (high priority) currently dismod_at.csv.predict will fail to generate a working ancestor.db file when the start_job_id argument points to a job wherein the parent location failed. This appears to be due to the check_log.py subroutine which starts checking the logs at the identified start_job rather than the root_node
  2. a request has been made to surface the integrand table from https://dismod-at.readthedocs.io/integrand_table.html in the at-cascade docs; their current location is somewhat esoteric and the information would be very valuable to end users of at-cascade
bradbell commented 1 month ago

Item 1: We need a test case, that is as simple and runs fast and demonstrates the bug mentioned in this iterm above. This confirms the problem and will become part of the regression testing after it is fixed.

Item 2: Who is the audience for the documentation that you are referring to here; .e.g., is it users of the csv interface, users of at_cascade, or developers of at_cascade ?

ntemiq commented 1 month ago

Item 1: I'll get working on a simple test case

Item 2: It is for users of the csv interface.

bradbell commented 1 month ago

It seens to me that what you really want is a separate buld of the documentation that is just for users of the csv interface and does not have links to the rest of the at_cascade or dismod_at or cppad_mixed documentation ?

bradbell commented 1 month ago

I think that we should use the special message type `at_cascade`` to make it easier to find which fits failed; see https://dismod-at.readthedocs.io/log_table.html#message-type https://at-cascade.readthedocs.io/add_log_entry.html

This was done; see changes to fit_one_job.py and check_log.py in https://github.com/bradbell/at_cascade/commit/3febb54c935af85a87c44cadc538c3bd4df780c88729afa

bradbell commented 1 month ago

I think that item 1 was fixed by the change to at_cascade/csv/pre_parallel.py in https://github.com/bradbell/at_cascade/commit/374d39f567c75aec533c96273449f9b6185ccbf7

It would be good if you had a test case to verify this.

ntemiq commented 1 month ago

@bradbell test for item 1 in linked PR above