Extract from db without prodigy

1

Have you considered making it possible to read saved annotations from DB without having to install prodigy?

Use case

I have a spacy project where I want to control the version of spacy myself. But I also want to be able to download the data annotated in prodigy - currently I can only do that by installing prodigy which then locks my spacy version (and other dependencies)

Workaround

My current workaround is that I have a copy of all my labelled data in other tables where I know how to serialize and deserialize but it just feels wrong (I have a prodigy recipe to do this....).

Looking in prodigy.compoonents.db I see the following dependencies that are not available open source

from ..util import PRODIGY_HOME, TASK_HASH_ATTR, INPUT_HASH_ATTR, log, registry
from ..util import ENV_VARS, get_config, convert_blob, get_display_name

If those were open source then I could just extract the db module from prodigy - especially convert_blob is crucial AFAIK.

Whats your thoughts on this?

2

Hi @nix411 ,

My suggestion is to interact with the database directly--if you're using the default db (SQLite), then you should be able to load it using any Python ORM library (like peewee or the built-in sqlite).