Update sqlite3.ts

[light-merlin-dark]

1n and 0n were causing error: RangeError: Loss of precision reading BigInt (0)

This resolved it:

else if (typeof value === "boolean") {
        return value ? 1 : 0;
    }

Was there a special intention to convert a boolean to a bigint? If not or if this was in error, it seems like this resolves and simplifies this part.

Thank you for all of your time and energy into this wonderful library.

🤍🙏🏼🌹⚡️✨

[penberg]

Hey @light-merlin-dark! Can you share the code snippet that was causing you the original error? I am not actually sure why we encode booleans with BigInt, but it's also part of the libSQL remote protocol ("hrana") so it's probably hard to change.

[light-merlin-dark]

Hey @penberg, good day! ☀️

For clarity, from the history of sqlite3.ts, it looks like the change was introduced in the commit: add more tests for numeric values.

I encountered the issue in my stack with bun and drizzle-orm. I did a push to a remote turso db and then upon reading a table I encountered the error above.

I made the change to revert it back to use normal integers and that resolved the error.

In further research with ChatGPT 4, it conveyed that the code to convert booleans as Bigints didn't seem standard unless it was for future-proofing or something like that.

All for clarity, and thanks again for the wonderful library to all who've contributed so far.

🤍🙏🏼🌹⚡️✨

[giovannibenussi]

Hey @light-merlin-dark, are you still having this issue? I just tested pushing a schema and then reading a table using bun and drizzle-orm and it worked well for me. Do you have a way to reproduce this issue to investigate it further?

[light-merlin-dark]

Hey, @giovannibenussi. Good day! ☀️

Regarding the issue, I got past it with my update to that part of the code and I currently don't have the context to revert back or check again.

It seems like something with a nature that can be determined in it's local context, whether to convert a boolean the previous way or a more common way and what support that may have across libraries. Of course, I think the most compatible and performant way is the way to go, and it seems like, with the nature of that specific function it's easy to determine if it serves it's purpose without introducing an error.

Thanks again for building such a wonderful library. I wish you and everyone here such a great day.

🎩🌹🌙🪽✨

[giovannibenussi]

Thanks! Sounds good to me. The current code looks good to me and your issue is solved so I'll close this PR.

[giovannibenussi]

Opening this ticket since we found a new use case for this.

Nmans01/lib-bool-reproduce is a simple reproducer project that uses Bun and Prisma. When run it, you see the same Loss of precision reading BigInt (1) error than this issue mentions.

The cause is the boolean inside the data that is being saved:

    data: {
        name: 'Jon',
        enabled: true,
    }

To confirm this, you can do the following:

• Go to the prisma/schema.prisma file and replace enabled Boolean by enabled Boolean? to make the argument optional.
• Run bun run prisma db push to update the schema on the database.
• Remove enabled: true from the data.

Since this is a fairly common use case, I think that we should merge this PR unless there's a strong reason to not do so.

Edit: this is being worked on this PR.