Implement high-precision 128-bit value types #2072
Conversation
Fix databento decoder logic
cjdsellers
changed the title
nautilus_trader/model/objects.pyx
Outdated
| @@ -67,8 +73,10 @@ PRICE_MIN = RUST_PRICE_MIN | |||
| MONEY_MAX = RUST_MONEY_MAX | |||
| MONEY_MIN = RUST_MONEY_MIN | |||
|
|
|||
| FIXED_PRECISION = RUST_FIXED_PRECISION | |||
| FIXED_SCALAR = RUST_FIXED_SCALAR | |||
| HIGH_PRECISION = PRECISION == 16 | |||
There was a problem hiding this comment.
It'll be a safer option to load this from the cdef HIGH_PRECISION value defined in model.pxd. Just in case the digits of precision change for high precision it'll very difficult to track down why this is failing.
There was a problem hiding this comment.
Also had a similar thought, had to add another constant generated at compile time, as HIGH_PRECISION was already occupied with the DEF, but now achieving what we were after:
HIGH_PRECISION = RUST_HIGH_PRECISION_MODE
Codecov ReportAttention: Patch coverage is 📢 Thoughts on this report? Let us know! |
| @@ -191,8 +186,9 @@ mod tests { | |||
| types::{Price, Quantity}, | |||
| }; | |||
|
|
|||
| #[cfg(feature = "high-precision")] // TODO: Add 64-bit precision version of test | |||
There was a problem hiding this comment.
We'll do another sweep of the tests once merged.
| ); | ||
| // TODO: Revisit calculations |
There was a problem hiding this comment.
This was outputting an unscaled raw quantity, we probably want the scaled quantity.
| @@ -15,7 +15,7 @@ | |||
|
|
|||
| //! Represents a medium of exchange in a specified denomination with a fixed decimal precision. | |||
| //! | |||
| //! Handles up to 9 decimals of precision. | |||
| //! Handles up to 16 decimals of precision. | |||
There was a problem hiding this comment.
Will need to tidy up the docs some more depending on precision mode.
| @@ -1201,13 +1200,13 @@ class PyCondition: | |||
| Condition.positive(value, param, ex_type) | |||
|
|
|||
| @staticmethod | |||
| def positive_int(int64_t value, str param, ex_type = None): | |||
| def positive_int(value: int, str param, ex_type = None): | |||
There was a problem hiding this comment.
Relaxed all of these C types, as they're coming from Python in non-performance critical function calls.
Avoids having to name the specific integer types in C.
Pull Request
high-precisionfeature flag to switch between lower and higher precision.See RFC.