[dccote]

First step of the redesign discussed during review of PR #85.

What changes

Every `Command` subclass — `TextCommand`, `MultilineTextCommand`, `DataCommand` — now carries the same four attribute names for the same four conceptual operations:

Role	Attribute	What TextCommand holds	What DataCommand holds
Encode outgoing request	`requestEncoder`	format string `"g r{register}\n"`	`DataEncoder('<clllc', fields, defaults)`
Decode incoming request (mock)	`requestDecoder`	regex `r'g r(?P...)'`	`DataDecoder('<xlllx', fields, prefix=b'M')`
Encode outgoing reply (mock)	`replyEncoder`	format string `"v {value}\r"`	`DataEncoder('<lllc', fields)`
Decode incoming reply	`replyDecoder`	regex `r'v\s(-?\d+)'`	`DataDecoder('<c', length=1)`

Two new frozen dataclasses (`DataEncoder`, `DataDecoder`) collapse the 12 previous attribute slots on `DataCommand` (`sendFormat`/`sendFields`/`sendDefaults`/`requestFormat`/`requestFields`/`prefix`/`responseFormat`/`responseFields`/`unpackingMask`/`replyDataLength`/`replyHexRegex`/`_prefix`) into 4 grouped slots.

Deleted attributes

• `alternatePattern` (TextCommand + MultilineTextCommand) — plumbed through `readMatchingGroups` but never set by any Command in the codebase. If a future caller needs alternation, write it into the `replyDecoder` regex.
• `replyHexRegex` (DataCommand) — only code path raised `NotImplementedError`. Dead.

Migration scope

All five device families' `commands = {...}` dicts migrated:

Device	Commands	New form's shape
`SutterDevice`	4 (binary, MP-285 protocol)	`DataEncoder` + `DataDecoder` triples
`IntellidriveDevice`	3 (text, named groups)	rename only
`IntegraDevice`	5 (text + multiline)	rename only
`EchoDevice`	3 (mixed)	rename + one `DataDecoder(length=...)`
`CoboltDevice`	11 (text)	rename only

Test fixtures in `testCommandRecognition.py` (38 tests), `testTableDrivenDebugPort.py` (3 fixtures + 12 tests), and `testCommunicationPorts.py` (5 tests) all rewritten for the new constructor signatures.

Before / after — SutterDevice MOVE

```python

Before

"MOVE": DataCommand(name="MOVE", prefix=b'M',
sendFormat='<clllc',
sendFields=('header', 'x', 'y', 'z', 'terminator'),
sendDefaults={'header': b'M', 'terminator': b'\r'},
requestFormat='<xlllx', requestFields=('x', 'y', 'z'),
replyDataLength=1, unpackingMask='<c'),

After

"MOVE": DataCommand(name="MOVE",
requestEncoder=DataEncoder('<clllc',
('header', 'x', 'y', 'z', 'terminator'),
{'header': b'M', 'terminator': b'\r'}),
requestDecoder=DataDecoder('<xlllx', ('x', 'y', 'z'), prefix=b'M'),
replyDecoder=DataDecoder('<c', length=1),
),
```

Internal renames

• `TextCommand.effectiveMatchPattern` (property) → `effectiveDecoder` (property)
• `TextCommand._autoMatchPattern` (was a property, never set state) → `_autoMatchPattern()` (now a method)

Both are private-ish; the only tests that touched them were updated.

Test plan

• Full suite: 261 passed, 203 skipped, 0 failed. Skips are hardware-gated and were also skipping before this commit.
• `testCommandRecognition.py` (36 unit tests): all pass with new attribute names.
• `testTableDrivenDebugPort.py` (12 dispatch tests): all pass with new fixture syntax.
• Mock-side device tests for Sutter / Intellidrive / Echo / Integra / Cobolt: all pass.
• Verify against a live Sutter / spectrometer when convenient.

What this PR does not address (intentional, deferred)

From the review of PR #85, the redesign concerns were:

1. ✅ TextCommand and DataCommand don't share a shape — done in this PR.
2. ⏭ Schema vs instance conflation — `Command` still holds runtime state (`reply`, `matchGroups`, `isSent`) on the same object as the immutable schema. Separating these is the next step.
3. ⏭ Mock hook contract weakness — `process_command(name, params, endPointIndex)` still receives `name` as a string and `params` as either tuple-or-dict. Tightening this is downstream.
4. ⏭ Error model inconsistency — TextCommand swallows + stores in `.exceptions`; DataCommand re-raises. Unchanged here.
5. ⏭ `endPoints` on Command — kept as-is per your call.

Relationship to PR #85

`unify-sendcommand` (PR #85) is unaffected by this PR — they touch different concerns. After this PR merges, PR #85 will need a rebase that picks up the new attribute names.

🤖 Generated with Claude Code