Thanks for the project, and a quick question about related work

[Balaxxe]

Hi! First, thank you for creating and maintaining OpenSQLAnywhere. I really appreciate the initiative, and I would love to give something back to the project if I can.

I recently committed some typed-row decoding work in my fork. The implementation was cross-checked against the independently developed, MIT-licensed DBReader project. Since the contribution policy refers to format knowledge coming from the contributor's own analysis, I wanted to check with you before linking anything.

If that provenance is acceptable, I would be happy to share an immutable commit link for review. If you would prefer not to review it under the project's clean-room policy, that is completely fine too.

Please also feel free to completely ignore this. There is no expectation that you respond or do anything with it. I mainly wanted to say thank you for the project and let you know that I appreciate the work you are doing.

[Balaxxe]

A summary, if it helps:

The fork now has a substantial extension to the page-store layer, including:

• Strict validation of already-materialized type-4 table pages
• Bounded directory and physical-record access
• Bounded physical row-segment parsing and caller-controlled continuation walking
• Conservative page-link parsing with range and cycle protection
• Schema-driven typed-row decoding for integer, numeric, date, date-time, character, text, boolean, nullable, and variable-width fields
• Explicit handling of row references, overflow states, unsupported types, ambiguous layouts, and malformed boundaries
• Partial-row decoding for callers that can prove only a safe prefix

The implementation remains generic to the SQL Anywhere storage layer. QuickBooks-specific schemas and accounting behavior stay in OpenQBW. The DBReader provenance and MIT license are recorded directly in the repository.

[Nathan-D-R]

Hi Balaxxe,

Sorry for the (really) slow reply. I should have sent confirmation that I saw it instead of sitting on it until now. I've been going back and forth on this for a couple of weeks, which is honestly a compliment to the work you described. I didn't want to fire off a quick no, and I didn't want to say yes and compromise the projects. Thanks for asking first instead of just opening a PR - that was really respectful of you. Replying to both of your issues here so it's in one place.

On OpenSQLAnywhere: yes, please send it. I went and looked at DBReader and I'm satisfied with it. MIT, and the Sybase unit says the format was worked out from sample databases, which is the same standard I hold this project to.

Three things before you open a PR:

1. Was the SQL Anywhere work done independently of your QuickBooks SDK use? QuickBooks sits on top of SQL Anywhere, so the two aren't automatically separate even though you kept the code separate. Answer at the level of where the knowledge came from, not what it turned out to be. Shorter is preferred in this case.
2. Confirm you never ran or validated against SAP SQL Anywhere itself.
3. The PR needs DBReader's actual MIT text and Sergey Bodrov's copyright line in NOTICE. Apache-2.0 takes MIT inbound fine, just need to satisfy the MIT license requirements.

Past that it's a normal PR and CONTRIBUTING has the details.

On Sigilweaver/OpenQBW#18: I have to pass, and please don't send the link. It isn't about the quality or anything. Once I've read something that was checked against Intuit's SDK I can't unread it, and I lose the ability to say this project was written independently.

What I'd genuinely value there is plain bug reports. What file you used, what you expected versus what you got. Just be careful to leave out the diagnosis - so no offsets, field meanings, or expected numbers that came out of QuickBooks. I know that's an odd thing to ask of someone who already knows the answer. Point me at what looks wrong and let me work out why.

Most of what I test against is the QuickBooks practice files from https://www.tlr-inc.com/quickbooks-practice-files/ - the student and textbook company files, mostly the Get Smart 2021 and 2022 bundles. Those are training data rather than anyone's real books, and they're free to download, so if you file reports against those I can reproduce directly. Their terms don't allow redistributing the files themselves, but they're easy enough to get from the source. If you use something different I would prefer if it's publicly accessible with permissive terms and not something you personally generated.

Either way, I'd love to credit you for anything that lands on either project!