Heads-up: `datalogic-rs` v5 is landing

[shankar-gpio]

Hi! Maintainer of datalogic-rs here. The Rust SDK contrib workspace pins datalogic-rs = "4.0.4", and we're about to publish v5.0.0 a coordinated breaking release across the Rust core and all bindings.

• v5 branch: https://github.com/GoPlasmatic/datalogic-rs/tree/v5
• Migration cookbook: https://github.com/GoPlasmatic/datalogic-rs/blob/v5/MIGRATION.md

The Rust renames are mechanical 1:1 search-and-replace covers ~90% of call sites. The 4.x line stays installable for anyone not ready to move.

New in v5: native bindings for other languages, all sharing the same Rust core. Might be relevant for the wider OpenFeature SDK ecosystem if flagd evaluators in other languages would benefit from a uniform JSONLogic engine with matching semantics:

• @goplasmatic/datalogic-node (Node, via napi-rs per-platform prebuilds)
• @goplasmatic/datalogic-wasm (browser / edge / Deno / Bun / Workers)
• datalogic-py (Python, via pyo3 - abi3-py310 wheels for Linux/macOS/Windows)
• datalogic-go (Go, over the C ABI)
• datalogic-c (C ABI via cbindgen)

Happy to open the migration PR against this repo if that's easier. Otherwise, any feedback on MIGRATION.md before we cut the release would be very welcome.

Thanks!

[beeme1mr]

Hey @shankar-gpio, thanks for the heads up and congratulations on the release. Will the native bindings support custom operators? We use a few that are feature flag-specific.

[shankar-gpio]

Thanks, we are yet to release the v5. Working with other repos to analyse the integration challenges before releasing.
Good catch, currently the bindings do not support custom operator. I'll add that before releasing.

[shankar-gpio]

Hey, quick follow-up on both of your questions — v5 is just about ready and both pieces have landed.

Custom operators in the bindings

All five v5 bindings now let you register your own operators in the host language. The callback contract is the same everywhere: it receives the operator's evaluated args as a JSON-array string and returns a JSON-value string.

• WASM (@goplasmatic/datalogic-wasm): new Engine({ customOperators: { name: fn } })
• Node native (@goplasmatic/datalogic-node): new Engine(opts, { name: fn })
• Python (datalogic-py): Engine(custom_operators={"name": callable})
• C ABI (bindings/c): datalogic_engine_builder_add_operator(builder, name, cb, user_data)
• Go: NewEngineBuilder().AddOperator(name, fn).Build()

Design notes and acceptance criteria are in GoPlasmatic/datalogic-rs#49, and there's a quick reference at bindings/BINDINGS.md.

fractional and sem_ver are built-in now

Since you mentioned flag-specific operators — instead of leaving these for every downstream to re-implement, we've shipped both as opt-in built-ins behind a flagd Cargo feature:

datalogic-rs = { version = "5", features = ["flagd"] }

It's off by default to keep the lean build lean. The implementations follow flagd's canonical Go evaluator closely — the test suites under tests/suites/flagd/ are direct ports of the upstream Go tests so the behavior stays in sync.

Two small things worth flagging while looking at this:

1. fractional here uses the same integer distribution math as the Go evaluator ((hash * total_weight) >> 32). The version currently in this repo uses a float-percentage form, and the two can pick different variants for inputs near a bucket boundary — so flag stickiness wouldn't quite match other flagd providers.
2. sem_ver here applies the four input normalizations from the flagd spec (strip v/V, pad partial versions, coerce numbers to string, ignore build metadata), which the current version in this repo doesn't.

Tracking issue: GoPlasmatic/datalogic-rs#50. Operator docs: docs/src/operators/flagd.md.

What this means here

Once v5 ships, the targeting/fractional.rs and targeting/semver.rs modules can be removed — flipping features = ["flagd"] on the datalogic-rs dep registers both automatically, and they'll behave identically to every other flagd provider. Happy to put up the migration PR alongside the v5 bump.

[jbovet]

Hi all, the workspace was bumped to datalogic-rs = "5.0.0" in #144 (863cd71), and the in-process evaluation works in #147