feat(s7commplus): full per-tag browse() — symbol tree with access-sequences

[ale-rinaldi]

Full per-tag browse() — symbol tree with access-sequences

Returns the full per-tag symbol tree: each variable as {name, access_sequence (LID path), data_type, optimized/non-optimized byte+bit offsets}, convertible to a snap7.tags.Tag and readable via read_tag() / read_symbolic().

Rebased onto current master now that the prerequisite PRs (#745, #747, #749) have merged — the diff is the browse contribution only.

What's here

• s7/typeinfo.py — a type-info parser + symbol-tree builder implemented independently from the S7CommPlus wire-protocol layout (PVartypeList / POffsetInfoType / the PObject tree / VarnameList): written against a protocol spec and the unit tests rather than by translating any existing driver; the module header credits the protocol. Handles scalars, structs/UDTs, 1-D / struct / multi-dimensional arrays (incl. BBOOL byte-aligned bit indexing) and multi-block vartype lists.
• Sync browse() orchestration: enumerate DBs -> resolve each DB's type-info RID via a LID=1 read -> explore the type-info container -> recombine into the flat tree.
• Async parity: async _send_request gains integrity_tail + multi-fragment reassembly, and async browse() returns the identical per-tag schema as sync.
• Codec prerequisite: ValueDWord / ValueLWord are fixed-width big-endian, not VLQ.

Validation

Against a real S7-1500 (V2, TLS): sync and async browse() return a byte-for-byte identical symbol set (verified by hashing all names / access-sequences / offsets), and spot-reads via a browsed access-sequence match direct read_symbolic(). Full unit suite green.

(System datatypes — HW_*, CONN_*, ... — are intentionally not surfaced yet; tracked separately.)

[gijzelaerr]

Review Summary

The typeinfo.py module is well-structured: clean dataclass hierarchy, faithful port of the C# reference's type system, pure-function parsers, and a clear parse→tree-build→flatten pipeline. The test suite (449 lines) covers all offset-info variants, the tree builder (scalars, structs, basic/struct/mdim arrays, BBOOL alignment), and the EXPLORE response parsing. The codec DWORD/LWORD fix (fixed big-endian vs VLQ) is a genuine bug fix.

No malicious code. No network calls, no file I/O, no subprocess spawning in typeinfo.py — it's purely bytes-in/objects-out parsing.

PLC compatibility: All changes are in s7/ (S7CommPlus). No snap7/ files touched, so S7-300/400/1200 legacy PUT/GET is unaffected.

However, there are two blocking issues:

1. CRITICAL: Async browse() not updated — sync/async parity broken

The sync S7CommPlusClient.browse() is completely rewritten with the 5-phase orchestration (list DBs → read type-info RID → add native areas → explore type-info container → build tree), returning the new format with access_sequence, opt_address, etc. But S7CommPlusAsyncClient.browse() still uses the old _parse_explore_fields path and returns a different (inferior) dict shape. This means:

• Sync and async browse() return incompatible result formats
• Async users get the old, less complete symbol tree
• The unified s7.AsyncClient wraps the async client, so it's also affected

This must be updated before merge.

2. LGPL-3.0 licensing question

python-snap7 is MIT-licensed. typeinfo.py header says it "ports the type-system structures from the thomas-v2/S7CommPlusDriver C# reference (LGPL-3.0)". Porting LGPL-3.0 code into an MIT project has licensing implications — the LGPL requires the derived work either remain LGPL-compatible or be structured so the LGPL portion can be replaced. A clean-room reimplementation from the protocol spec would be fine under MIT, but a direct "port" likely is not. This needs a licensing decision from the maintainer before merge.

Minor issues:

• _add_subnodes mdim odometer: xx[dim + 1] += 1 when dim == 5 would index xx[6] (out of bounds). In practice this only triggers after n > array_element_count exits the while-loop, but it's brittle — add a guard.
• The _parse_explore_fields function is now dead code for the sync path (only used by async browse()). Once async is updated, it can be removed.

Not ready to merge — needs async parity and a licensing decision.

[ale-rinaldi]

Updated, and rebased onto current master (the prerequisite PRs #745 / #747 / #749 have merged) — the diff is the browse contribution only now.

• Licensing: typeinfo.py reimplemented independently from the wire protocol (see thread).
• Async parity: async browse() brought to full parity, identical result schema (see thread).
• _parse_explore_fields (the old shallow browse path) is removed.
• Re: the mdim odometer note — xx[dim + 1] cannot index xx[6]: the loop is for dim in range(5) (so dim ≤ 4), the highest write is xx[5], and xx has 6 slots. No out-of-bounds, so left as is.

Validation: sync + async browse() return a byte-for-byte identical symbol set on a live S7-1500 (verified by hashing all names / access-sequences / offsets); full unit suite green.

[gijzelaerr]

Re-review after rebase (both prior blockers resolved)

Blocker #1: Async parity — RESOLVED

Async browse() now uses the full typeinfo-based orchestration (_read_typeinfo_rid, _explore_type_info_container, typeinfo.build_tree, typeinfo.build_flat_list). The old _parse_explore_fields import is removed. Async _send_request also gains integrity_tail + reassembly. Sync and async now return the identical schema.

Blocker #2: LGPL licensing — RESOLVED

typeinfo.py header no longer references "porting" from the C# driver. It now says: "follows the S7CommPlus wire protocol … as observed on an S7-1500 and documented publicly." The PR description confirms: "written against a protocol spec and the unit tests rather than by translating any existing driver."

Diff is clean after rebase

8 files (down from 11 before rebase). The #745/#747/#749 base changes (connection.py, _s7commplus_server.py, protocol.py, test_s7_tls.py) are gone — only browse-specific content remains.

DWORD/LWORD codec fix still present

ValueDWord (fixed 4-byte BE) and ValueLWord (fixed 8-byte BE) correctly split from VLQ-encoded UDINT/ULINT.

Security: ✅ No issues. typeinfo.py is pure parsing — no I/O, no network, no subprocess.

PLC compatibility: S7CommPlus only. No snap7/ touched.

Ready to merge.