Precision issue with lossless-json

[ThallesP]

Hey, it's great to hear that you now have an SDK. I maintained a rather rudimentary SDK at https://github.com/ThallesP/coingecko-node-sdk, and we're thinking of migrating to this one.

However, I have a question: how will CoinGecko handle precision issues? JavaScript is not a perfect language, and as soon as we parse the request from CoinGecko, that will cause number precision issues. In my SDK, I used lossless-json which parses the numbers as strings. Will the CoinGecko SDK follow something similar?

[cg-eesuhn]

Hi @ThallesP, thanks for the question.

At the moment, our SDK follows standard JSON parsing behavior where numeric values are parsed as number (IEEE-754 float64), which is consistent with how most JSON libraries behave.

We don't currently use a lossless parser, and if you need higher precision than float64, the recommended approach is to treat those fields as strings at the application layer or use a custom parser on top of the response.

We will close this issue for now, and appreciate the feedback from your end :))

[ThallesP]

Hi @ThallesP, thanks for the question.

At the moment, our SDK follows standard JSON parsing behavior where numeric values are parsed as number (IEEE-754 float64), which is consistent with how most JSON libraries behave.

We don't currently use a lossless parser, and if you need higher precision than float64, the recommended approach is to treat those fields as strings at the application layer or use a custom parser on top of the response.

We will close this issue for now, and appreciate the feedback from your end :))

Hey @cg-eesuhn, I’m just taking a look at this now, so in theory, I can’t use the SDK you provide? We treat numbers as strings at our application layer, but the problem is that the SDK itself parses them incorrectly :/

Any docs on how to plug a lossless parser in here? That might even affect the data types returned

[cg-eesuhn]

@ThallesP, thanks for the follow up
Would you mind sharing the following so we can better narrow down the issue:

1. The exact function/endpoint call you are making (including parameters)
2. The response you are getting back (with the incorrectly parsed values)
3. The expected response (what the correct values should be)

That way we can pinpoint exactly where the precision loss is happening and figure out the best path forward.