#udt #macro-derive #default #default-value #parser #rows #compatibility

macro scylla-macros-flex

This fork of scylla-macros provide alternative parsers with improved compatibility when reading rows & UDTs

15 stable releases

1.5.5 Oct 2, 2024
1.5.4 Dec 30, 2023
1.5.3 Oct 7, 2023
1.5.2 Apr 30, 2023
1.0.0 Apr 23, 2022

#1101 in Database interfaces

MIT/Apache

28KB
431 lines

scylla-macros-flex

This crate defines two derive macros:

  • FromRowFlex:

    • Any NULL value is converted to the default value (default for option is none, don't use this derive macro if one of the fields doesn't implement default).
  • FromUserTypeFlex:

    • A field not defined in the struct will be ignore without throwing an error. Warning: this could lead to a dataloss if the udt is overwritten with this partial udt.
    • Any NULL value is converted to the default value (default for option is none, don't use this derive macro if one of the fields doesn't implement default).

serde_json support:

Allows you to use arbitrary structs serialized with serde_json, stored as text in Scylla/Cassandra.

They have approximately the same behavior as UDT.

Requires "serde_json" feature.

Derive Macros:

- FromJson
- IntoJson

Pros:

- Way more flexible than CQL **FROZEN** UDTs thanks to Rust enums
- Schema can be changed a lot without any issue.

Cons:

- Can take more space than UDT
- Can be slower than UDTs

rmp-serde support:

Allows you to use arbitrary structs serialized with MessagePack, stored as Blobs in Scylla/Cassandra.

They have approximately the same behavior as UDT.

Requires "rmp-serde" feature.

Derive Macros:

- FromMessagePack
- IntoMessagePack

Pros:

- Way more flexible than CQL **FROZEN** UDTs thanks to Rust enums
- Schema can be changed a lot without any issue.
- Most of the time uses less space than UDT

Cons:

- Can be slower than UDTs

speedy support:

Allows you to use arbitrary structs serialized with speedy, stored as Blobs in Scylla/Cassandra.

Use this only if you won't change your schema later. They have approximately the same behavior as UDT.

Requires "speedy" feature.

Derive Macros:

- FromSpeedy
- IntoSpeedy

Pros:

- Way more flexible than CQL **FROZEN** UDTs thanks to Rust enums
- Less storage required than UDTs because udt include fields names & udt names
- Faster than UDT (because speedy is quite fast)

Cons:

- You can't update a field in an atomic manner (like frozen UDTs)
- Changing your schema (eg. adding a field to a struct inside a Vec) will break everything.

Dependencies

~1.5MB
~37K SLoC