feat: zlib preset dictionary decompressor for S7CommPlus blobs

[gijzelaerr]

Summary

• Adds the 19 known Siemens zlib preset dictionaries (24KB total) extracted from thomas-v2/S7CommPlusDriver (LGPL-3.0)
• Adds blob_decompressor.py with decompress_blob() that auto-selects the correct dictionary by Adler-32 checksum
• Replaces the partial oracle-based dictionary reconstruction from S7-1200 FW V4.5: I/Q/M symbolic tag browse via oracle-reconstructed preset dictionary (33/40 tags complete) #757 with complete, verified dictionaries

The key dictionary for #757 is IntfDescTag (Adler-32 0xce9b821b, 1178 bytes) — used for I/Q/M area tag definitions on V3 PLCs (S7-1200 FW V4.5+). With the complete dictionary, decompression produces clean XML that can be parsed with standard tools instead of regex extraction from garbled output.

Addresses #757.

Test plan

• Verify all 19 dictionary Adler-32 checksums match
• Decompress standard (no-dict) and preset-dict blobs
• Unknown dictionary raises clear error
• find_and_decompress() scans raw payloads for zlib streams
• @tommasofaedo: test against S7-1200 FW V4.5 I/Q/M EXPLORE responses

🤖 Generated with Claude Code

[tommasofaedo]

Independently validated find_and_decompress() from this branch against a live S7-1200 G2 (FW V4.1) — it works out of the box. 🎉

I no longer have the G1 FW V4.5 unit from #757, so all testing here is on the G2 only — but it turns out the G2 declares the exact same preset dictionary. All three EXPLORE areas carry a CMF=0x78 FLG=0x7d stream with dict Adler-32 0xce9b821b (IntfDescTag), and find_and_decompress(full_explore_payload) returns clean, zero-null XML:

<?xml version="1.0" encoding="utf-8"?><IdentContainer>
<Ident Name="mtag_word" Scope="Global" LID="13"><SimpleType>Word</SimpleType><Access SubClass="SimpleAccess"><SimpleAccess ByteNumber="102" Width="Word" Range="Memory" /></Access></Ident>
...
</IdentContainer>

Follow-up wiring (the part you flagged as TODO)

Since the XML is clean, the tag layer is tiny. Mapping Range+Width+ByteNumber+BitNumber to TIA syntax:

import xml.etree.ElementTree as ET
from s7commplus.blob_decompressor import find_and_decompress

_RANGE = {"Input": "I", "Output": "Q", "Memory": "M"}
_WIDTH = {"Byte": "B", "Word": "W", "DWord": "D", "LWord": "L"}

def _address(rng, width, byte_no, bit_no):
    letter = _RANGE.get(rng, "?")
    if width == "Bit":
        return f"%{letter}{byte_no}.{bit_no or 0}"
    return f"%{letter}{_WIDTH.get(width, '')}{byte_no}"

def parse_ident_container(xml_text):
    tags = []
    for ident in ET.fromstring(xml_text).findall("Ident"):
        acc = ident.find("./Access/SimpleAccess")
        if acc is None:
            continue
        byte_no = int(acc.get("ByteNumber", "0"))
        bit_raw = acc.get("BitNumber")
        tags.append({
            "name": ident.get("Name"),
            "type": ident.findtext("SimpleType"),
            "lid": int(ident.get("LID", "0")),
            "address": _address(acc.get("Range"), acc.get("Width"),
                                byte_no, int(bit_raw) if bit_raw else None),
        })
    return tags

Result against the running G2 (verified against TIA Portal):

Name	Type	Address	LID
input_1	Bool	%I0.0	10
output_1	Bool	%Q0.1	9
merker_1	Bool	%M0.2	9
mtag_byte	Byte	%MB100	11
mtag_word	Word	%MW102	13
mtag_dword	DWord	%MD104	14
mtag_bool	Bool	%M108.0	16

This makes the whole oracle pipeline from #757 (_build_fdict / regex _extract_tags / symbolic-READ width workaround) obsolete — the <SimpleType> now carries the real M widths directly.

I'm opening a small stacked PR on top of this branch that adds a non-invasive tag_browser.py (the parser above + an EXPLORE-per-area helper), so you can wire it into browse() however you prefer — or just fold it into this PR. Whatever's least noisy for you. Only G2-validated on my end, flagging that honestly. Thanks again for the complete dictionaries.