[DomCR]

Description

Transfer PR for:

• feat: read native Tianzheng text objects #1275

Tasks done:

• Code cleanup
• Refactor and standardize the reader.
• Stablish the data structure of the TEXT entities as a placeholder.

[DomCR]

Tianzheng (TArch) TCH_TEXT / TCH_MTEXT – Findings

1. AC1021 sample (issue-1131_TZ.dwg)

Object stream (after readCommonEntityData)

• Fixed 82-bit header: 0A 11 90 55 05 90 39 03 81 04 + 2 bits 01
  • Identical for TCH_TEXT (828, 829, 935) and TCH_MTEXT (949), so it's a shared Tianzheng header.
  • Reading it as BitLong, BitShort or BitDouble gives no sensible values. Meaning unknown.
• Insertion point: 3 BitDoubles in the order Y, 0, X at offset +82 (all 12 entities).

String stream (numbers are TgDataEncrypt ciphertext)

Field	TCH_TEXT (11)	TCH_MTEXT (949)
ClassId	{F8F0F12F-…}	{DF708B01-…}
Scale	100	100
unknown double / int	0 / 0	0 / 0
TDBXDATAMAP_BLOCK_BEGIN_ … END_	T20d1..T20d4	T20d1..T20d4
StyleId	{B0713E6B-…}	same
unknown0	0	0
Value	text	multiline (\r\n)
Height	3.5	3.5
rotation?	0	0
unknown1	7	10
unknown2	0	0
Width	–	200
LineSpacingFactor	–	1.4

• Both classes use the same layout up to unknown2. MText adds Width and LineSpacingFactor.
• unknown1 depends on the class: alignment code or version number per class.
• Rotation and the other unknowns are 0 in every sample, so they can't be identified yet.

2. AC1027 sample (TCH_TEXT handle 19)

How the DWG reader handles the two versions

	AC1021	AC1027 (R2010+)
String stream located	updateHandleReader() using the object-size (RL) field	getEntityType() using the MC handle-stream size
Common entity data	R2007 fields	+ visual style bits

Standard entities read correctly in both versions, so the common data is aligned.

Layout of handle 19

• Object data after the common entity data: 312 bits

  07 34 E7 D6 CE D5 6A 18 50 20 08 08 FD 19 E1 C4 EC 20 21 42
  1E 22 8F E1 20 D4 30 90 60 00 00 00 00 00 00 51 02 05 00
  

• String stream: only TDBXDATAMAP_BLOCK_BEGIN_ and TDBXDATAMAP_BLOCK_END_ (empty DataMap).
• 19 bits remain before the handle section.
• Different header, no point at +82, no ClassId, ciphertext, StyleId or text value.
• The binary-looking strings and the OverflowException from an earlier run came from reading past the end of the string stream.

3. TArch version issue

	AC1021 file	AC1027 file
Header	0A 11 90 55 …	07 34 E7 D6 …
Point	Y/0/X at +82	not found
String preamble	ClassId, Scale, …	none
Text content	in the string stream	none

Possible causes (not confirmed):

1. Different TArch versions wrote each file. The format depends on the TArch release, and the header acts as a version marker.
2. TArch switches layout by target DWG version in its own dwgOutFields.
3. Handle 19 isn't a drawn text. It's early in the file, with an empty DataMap and no content, so it may be a template or default object.

The reader currently supports only the AC1021 TArch layout if it's aligned with the dwg version and not the TArch app version. With other files it produces wrong values or reads past the end of the string stream.

This may be a critical issue when reading the entities, if TArch has multiple formats for the different version that makes the job of reading the value much harder and the stability cannot be granted.