Skip to content

feat: add ROWID column type (type# 69) decoding#319

Open
rophy wants to merge 1 commit intobersler:masterfrom
rophy:fix/rowid-decode
Open

feat: add ROWID column type (type# 69) decoding#319
rophy wants to merge 1 commit intobersler:masterfrom
rophy:fix/rowid-decode

Conversation

@rophy
Copy link
Copy Markdown

@rophy rophy commented Mar 20, 2026

Summary

ROWID columns (Oracle type# 69) are currently unsupported — their values are
silently dropped from the output. This adds decoding support.

Details

ROWID in redo logs is stored as 10 bytes big-endian:

  • dataObj (4 bytes) + DBA (4 bytes, afn<<22|block) + slot (2 bytes)

Output uses Oracle's standard base64 ROWID format (e.g. AAASnuAAMAAAAKtAAA),
matching what SELECT ROWID FROM ... returns.

UROWID (type 208) is unchanged — it keeps the existing hex format since UROWID
can store variable-length logical ROWIDs from index-organized tables where base64
encoding would be incorrect.

Changes

  • SysCol.h: add ROWID = 69 to COLTYPE enum
  • Builder.cpp: decode 10-byte ROWID and output as base64 string
  • BuilderJson.h: add ROWID to known types filter + schema metadata
  • BuilderProtobuf.h: add ROWID to known types filter + schema metadata

Verification

Tested against Oracle Free 23ai — ROWID values match LogMiner's SQL_REDO output.
Format verified by hex-dumping raw redo log bytes and matching against LogMiner's
base64 ROWID.

ROWID columns were unsupported — values were silently dropped. Add
decoding from the 10-byte big-endian redo format (dataObj + DBA + slot)
and output as Oracle's standard base64 ROWID string.

UROWID (type 208) is unchanged — it keeps the existing hex format since
UROWID can store variable-length logical ROWIDs from IOTs.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant