ARAX seems to ignore the path locations for database file specified in `config_dbs.json`

[saramsey]

It seems that node_synonymizer.py, when it is attempting to find the Node Synonymizer Sqlite file, looks in config_dbs.json to get the filename, but ignores the path provided in config_dbs.json, even though the config file appears to provide a full path location:

"node_synonymizer": "/translator/data/orangeboard/databases/KG2.10.2/node_synonymizer_v1.0_KG2.10.2.sqlite",

After some testing, and after reviewing the code in node_synonymizer.py, I am pretty confident in saying that the node synonymizer just pulls the filename (node_synonymizer_v1.0_KG2.10.2.sqlite) of the sqlite file from the config file and then assumes that the sqlite file is located in the same directory as node_synonymizer.py. Why is that? Why do we allow the user to specify a sqlite file location in config_dbs.json but then ignore the configured path location, when looking for the synonymizer file?

For testing ARAX on a local dev machine, it would sure be nice if I could configure alternate locations for the dbs, using the config_dbs.json file. But at least as concerns node synonymizer, it appears that I cannot do this without editing python code or creating a sym link. But in that case, why do we have a sym link? Why doesn't the module just use the configured directory location?

[saramsey]

I suspect this evolved because inside the Docker container, for whatever reason, the directory of, say, the node synonymizer sqlite file is not /translator/data/orangeboard... but in something like /mnt/data/orangeboard/.... Anyhow, some day, we should try to fix this. But it looks like it might be a bit tricky to do it without causing breakage somewhere in the codebase.

[edeutsch]

This might be an artifact of when I wrote the first NodeSynonymizer 8 years ago, where the sqlite file was in the same folder as the code. and then later when we started the whole /databases thing, we made a symlink from the code dir file to the /databases file, and the silly thing I did 8 years has been carefully preserved ever since? Even in Amy's re-write?

I suspect that this legacy weirdness could be corrected. The ARAX_database_manager might also need to be updated (this creates the sym link) if this is all to be tidied up.

[saramsey]

Creating a branch issue-2626 for this work

[saramsey]

I have committed a patch for node_synonymizer.py, so that it will respect the full path to the node synonymizer sqlite file, if it is specified in the config file. Now, for this to work inside the docker container, I think there will need to be a single sym link /mnt/data that points to /translator/data or some such. Perhaps we are already creating it as a part of the image build, I would have to check. Anyhow, this branch is not ready for testing yet. But this patch at least illustrates what I had in mind. Along the way, I updated the path manipulation code so that it is portable across POSIX-compliant systems.

[saramsey]

Or, perhaps we could just set up the docker container so that /translator inside the docker container is the same as /translator outside. Maybe?