"small", "medium", and "large" IDL table attributes might be useful

[Galen Charlton]

It would be useful if the IDL could record whether a database table (or view) is typically expected to be "small", "medium", or "large".

The immediate impetus for this thought is work in Angular comboboxes that are meant to retrieve rows from a linked table. Knowing that the underlying table is small enough that it can be safely fetched in its entirety would allow pre-populating the entire dropdown. While a dev can do this right now on a case-by-case basis, decorating the IDL with cardinality values could be useful automation.

Conversely, marking a table as "large" could be used to signal PCRUD that it shouldn't honor requests to fetch an entire table atomically (or at least, not without making the client pass some sort of yes_i_really_mean_this flag).

[Blake GH (bmagic)]

Galen,

I like the idea! It makes me wonder if this information could be generated automatically. Some sort of automated table count(*) scan that would find the highest number of rows for all tables in all schemas and then calculate percentages for each table compared to the highest number?

small = < 50%
medium = > 50% < 80%
large = > 80%

I guess the issue is that pcrud needs this information to be in fm_IDL.xml at compile time? Is there a way around that? Like a dedicated table to hold this info, hooked up to an openSRF service? Perhaps autogen.sh -u would re-populate. Just getting the idea ball started.

[Bill Erickson (berick)]

Potentially related bug #1858233.