[DISCUSSION] Stuff to drop…

[ntninja]

While looking through the codebase I found several abstractions that don't appear to do anything other than adding code. Items on the following list should be resolved by a decision to either keep or drop them. Tricky cases may also have their own issues referenced from there.

• datastore.core.query.* and maybe others from that module ([UNDECIDED])
• datastore.core.serialize.Serializer and subclasses (see issue [DISCUSSION] Merge datastore.core.serialize and ipfshttpclient.encoding? #4)

[ntninja]

Considering that we read and write binary streams from our backend implementations, I don't see any reason for keeping the whole Query interface at all. AFAICT this was included to create a cross-backend language for querying things like NoSQL-databases (MongoDB). Since we don't need that for py-ipfs (datastructures are decoded using IPLD from binary blobs), I suggest we drop this code entirely and use IPLD, where it makes sense, instead.

[ntninja]

I'd also like to nominate some stuff from core/basic.py for deletion:

• NullDatastore as it clearly violates the library's constraints (use DictDatastore instead for testing)
• InterfaceMappingDatastore won't be useful anymore as we switch to making the library async (Make this library async-friendly #8)
• CacheShimDatastore is redundant when we have TieredDatastore + DictDatastore

Also all the remaining datastore “shims” are likely better placed into extra modules instead of cluttering core/basic.py.

(Deleting code is so much fun! 😁)

[ntninja]

I've dropped InterfaceMappingDatastore from my codebase as I don't see how it will be useful in an async/await library.

[AliabbasMerchant]

@Alexander255 Yes. Lets delete all the things you mentioned above.

[ntninja]

#12 drops all of InterfaceMappingDatastore, CacheShimDatastore & SymlinkDatastore

[ntninja]

Summary: serialize is keep (see issue #4); query is keep partially (see issue #14)