v2.3.0

Date: 2026-07-06
Previous release: v2.2.4

Summary

Adds typed codec decoding to the client: a new client decode command reads holding/input registers and decodes them into typed values (integers, floats, strings, BCD, time, IP, MAC, …) using the github.com/otfabric/go-modbus/codec registry, plus a client codecs command to discover available codecs. As part of this, the ad-hoc --ascii and --byteswap flags are removed from client read, which now focuses on raw capture and MCAP output. This is a breaking change for anyone using those two read flags.

Highlights

• client decode (new): Reads holding (FC3) or input (FC4) registers and decodes them with a single --codec (e.g. float32/layout:1234, ascii/registers:8, uint32/layout:4321, ip_addr, datetime_iec870_utc). The codec sets the per-value register width; omit --count to decode one value, or pass a multiple of the codec width to decode an array of values. Output honors --format text|json|table; JSON emits a typed values array. No MCAP side effect—decode is a read-and-interpret command.
• client codecs (new): Lists every registered codec (id, name, family, value kind, registers, bytes), with optional --family, --value-kind, and --registers filters and --format text|json|table. Mirrors the codec catalog exposed by the Modbus simulator server.
• Codecs source: The codec registry ships inside the already-depended-on github.com/otfabric/go-modbus (v1.1.0) under the codec subpackage—no new dependency and no version bump were required.
• Completion: New --codec flag supports tab completion (registered codec IDs); --function completion is reused on decode.
• Config / env: New MODBUSCTL_CODEC environment variable maps to --codec; MODBUSCTL_COUNT gains decode semantics (a multiple of the codec width; 0 = one value).
• client read (breaking): Removed --ascii / MODBUSCTL_ASCII and --byteswap / MODBUSCTL_BYTESWAP. The read JSON DTO drops pre_swap_hex, bytes_swapped, and ascii_decoded; read still captures raw bytes (raw_data_hex) and writes MCAP.

Upgrading

• Replace read --ascii: Use client decode --codec ascii/registers:N (or ascii_fixed / ascii_reverse / utf16be / utf16le variants) for text decoding.
• Replace read --byteswap: Use a layout-aware numeric codec via client decode (e.g. --codec uint32/layout:2143) instead of swapping bytes on a raw read.
• Scripts parsing read --format json: Stop relying on pre_swap_hex, bytes_swapped, and ascii_decoded; move typed interpretation to client decode and read its values array.
• Completion users: Regenerate and re-source your shell completion script so the new decode / codecs commands and --codec candidates are available.

Release workflow

• Unchanged from v2.2.4: reusable go-binary-release workflow and link-time version metadata.