Periph table fallback on TableChain for experimenter summary

[CBroz1]

Description

Thanks, @MichaelCoulter for submitting #1031

Previously, the SpyglassMixin delete process inherited banning of peripheral tables behavior, which is useful in the case of long-distance restrictions, and failed to fully inherit ignoring of non-spyglass tables. This cropped up for both (a) delete permission via the session table, and (b) downstream descendants for deleting part-masters.

I address this by giving TableChain a new fallback for the path between two nodes that includes peripheral tables, if no connection was previously found. This allows tables like IntervalLinearizedPosition to find a session connection, and ignores custom tables when checking downstream for part-master deletes.

Checklist:

• No. This PR should be accompanied by a release: (yes/no/unsure)
• N/a. If release, I have updated the CITATION.cff
• No. This PR makes edits to table definitions: (yes/no)
• N/a. If table edits, I have included an alter snippet for release notes.
• No. If this PR makes changes to position, I ran the relevant tests locally.
• I have updated the CHANGELOG.md with PR number and description.
• N/a. I have added/edited docs/notebooks to reflect the changes

[edeno]

@MichaelCoulter did this solution work for you?

[samuelbray32]

Was trying to test with data, but ran into issues from legacy tables in our database (#976)

[CBroz1]

Was trying to test with data, but ran into issues from legacy tables in our database (#976)

I'll see if I can find a systematic way to get rid of the empty legacy tables

[samuelbray32]

Thanks @CBroz1, as a note, the blocking ones I was hitting had small (~1 entries) likely used in development, but not definitively named as such

[samuelbray32]

@CBroz1 , wanted to know if you think this is a reasonable feature to include in this PR: I found a way to load all spyglass shared schemas such that their nodes are in the dependency graph.

import datajoint as dj
from spyglass.utils.database_settings import SHARED_MODULES

def load_spyglass_shared_schemas():
    # Get a list of all shared schemas in spyglass
    keys = dj.conn().query(
        """
                    SELECT
                        concat('`', table_schema, '`.`', table_name, '`') as tab, column_name
                    FROM information_schema.key_column_usage
                    WHERE table_name not LIKE "~%%" AND constraint_name="PRIMARY"
                    """
    )

    shared_schema = []
    for key in keys:
        schema_name = key[0].split(".")[0].strip("`")
        schema_prefix = schema_name.split("_")[0]
        if schema_prefix in SHARED_MODULES and schema_name not in shared_schema:
            shared_schema.append(schema_name)
            
    # Load the dependencies for all shared schemas
    for s in shared_schema:
        dj.schema(s).connection.dependencies.load()

    return shared_schema

load_spyglass_shared_schemas()

• Running this before constructing the RestrGraph allowed things to run through without importing specific shared tables. Overhead on run time was about ~5 seconds for this function, but it saved time in the delete call since dependencies were already loaded.

• As implemented this looks for schema with prefixes contained in SHARED_MODULES. Could extend to allow users to import schemas with their prefix into the graph easily

• Running this here could be an simple way to reduce need for imports

[samuelbray32]

Could include comment above, otherwise looks and runs fine