refactor: everything #10

nberlette · 2025-07-19T15:02:42Z

chore: move trig/const/rand tests to src dir
refactor(ieee754): abstract out common parts of float16/32/64 modules
refactor(float16): use ieee754 abstractions for float16
feat(float16): add float16/constants submodule
feat(float16): add float16/guards submodule
refactor(float32): use ieee754 abstractions for float32 ops
feat(float32): add float32/constants submodule
feat(float32): add float32/guards submodule

Copilot

Pull Request Overview

This PR refactors the IEEE-754 floating-point implementation by extracting common functionality into a shared abstraction layer. The changes consolidate duplicate code between float16 and float32 modules while adding new constants and type guard submodules.

Introduces a shared IEEE-754 abstraction module to eliminate code duplication
Adds constants and guards submodules for both float16 and float32 formats
Moves test files from root to src directory for better organization

Reviewed Changes

Copilot reviewed 14 out of 15 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
src/random.test.ts	Updates import path to relative reference
src/internal/ieee754.ts	New shared IEEE-754 encoding/decoding abstraction
src/float32/round.ts	Minor comment formatting improvement
src/float32/mod.ts	Adds exports for new constants and guards submodules
src/float32/guards.ts	New type guard functions for float32 validation
src/float32/encode.ts	Refactored to use shared IEEE-754 abstraction
src/float32/decode.ts	Refactored to use shared IEEE-754 abstraction
src/float32/constants.ts	New constants module for float32 format specifications
src/float16/mod.ts	Adds exports for new constants and guards submodules
src/float16/guards.ts	New type guard functions for float16 validation
src/float16/encode.ts	Refactored to use shared IEEE-754 abstraction
src/float16/decode.ts	Refactored to use shared IEEE-754 abstraction
src/float16/constants.ts	New constants module for float16 format specifications
src/constants.test.ts	Updates import path to relative reference

Copilot · 2025-07-19T15:05:25Z

src/internal/ieee754.ts

@@ -0,0 +1,227 @@
+// src/float/common.ts


The comment references an incorrect file path. It should reference 'src/internal/ieee754.ts' to match the actual file location.

Suggested change

// src/float/common.ts

// src/internal/ieee754.ts

Copilot · 2025-07-19T15:05:25Z

src/internal/ieee754.ts

+  nan: 0x7FFF8000000000000000000000000000n,
+  positive_infinity: 0x7FFF0000000000000000000000000000n,
+  negative_infinity: 0xFFFF0000000000000000000000000000n,
+  negative_zero: 0x8000000000000000n,


The negative_zero value for float128 appears to be incorrect. It should be 0x80000000000000000000000000000000n (32 hex digits) to match the 128-bit format, not 0x8000000000000000n (16 hex digits) which is for 64-bit.

Suggested change

negative_zero: 0x8000000000000000n,

negative_zero: 0x80000000000000000000000000000000n,

Copilot · 2025-07-19T15:05:26Z

src/float32/constants.ts

+ */
+
+/**
+ * The number of bits used for the exponent in the IEEE-754 half-precision


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* The number of bits used for the exponent in the IEEE-754 half-precision

* The number of bits used for the exponent in the IEEE-754 single-precision

Copilot · 2025-07-19T15:05:26Z

src/float32/constants.ts

+export const FLOAT32_EXPONENT_BITS = 8;
+
+/**
+ * The number of bits used for the mantissa in the IEEE-754 half-precision


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* The number of bits used for the mantissa in the IEEE-754 half-precision

* The number of bits used for the mantissa in the IEEE-754 single-precision

Copilot · 2025-07-19T15:05:26Z

src/float32/constants.ts

+export const FLOAT32_MANTISSA_BITS = 23;
+
+/**
+ * The exponent bias used in the IEEE-754 half-precision (or "binary32") format.


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* The exponent bias used in the IEEE-754 half-precision (or "binary32") format.

* The exponent bias used in the IEEE-754 single-precision (or "binary32") format.

Copilot · 2025-07-19T15:05:26Z

src/float32/constants.ts

+export const FLOAT32_EXPONENT_BIAS = 127;
+
+/**
+ * Well-known bit pattern representing `NaN` in the IEEE-754 half-precision (or


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* Well-known bit pattern representing `NaN` in the IEEE-754 half-precision (or

* Well-known bit pattern representing `NaN` in the IEEE-754 single-precision (or

Copilot · 2025-07-19T15:05:27Z

src/float32/constants.ts

+
+/**
+ * Well-known bit pattern representing `+Infinity` (positive infinity) in the
+ * IEEE-754 half-precision (or "binary32") format.


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* IEEE-754 half-precision (or "binary32") format.

* IEEE-754 single-precision (or "binary32") format.

Copilot · 2025-07-19T15:05:27Z

src/float32/constants.ts

+
+/**
+ * Well-known bit pattern representing `-Infinity` (negative infinity) in the
+ * IEEE-754 half-precision (or "binary32") format.


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* IEEE-754 half-precision (or "binary32") format.

* IEEE-754 single-precision (or "binary32") format.

Copilot · 2025-07-19T15:05:27Z

src/float32/constants.ts

+
+/**
+ * Well-known bit pattern representing `-0` (negative zero) in the IEEE-754
+ * half-precision (or "binary32") format.


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* half-precision (or "binary32") format.

* single-precision (or "binary32") format.

Copilot · 2025-07-19T15:05:27Z

src/float32/constants.ts

+
+/**
+ * Well-known bit pattern representing `+0` (positive zero) in the IEEE-754
+ * half-precision (or "binary32") format.


The comment incorrectly refers to 'half-precision' when describing float32 constants. It should refer to 'single-precision' instead.

Suggested change

* half-precision (or "binary32") format.

* single-precision (or "binary32") format.

nberlette added 8 commits April 22, 2025 04:50

chore: move trig/const/rand tests to src dir

646b748

refactor(ieee754): abstract out common parts of float16/32/64 modules

8573089

refactor(float16): use ieee754 abstractions for float16

0a9f46d

feat(float16): add float16/constants submodule

1a3a0be

feat(float16): add float16/guards submodule

2757b8c

refactor(float32): use ieee754 abstractions for float32 ops

7f4b40a

feat(float32): add float32/constants submodule

9728995

feat(float32): add float32/guards submodule

5888c28

nberlette added documentation Improvements or additions to documentation enhancement New feature or request labels Jul 19, 2025

Copilot AI review requested due to automatic review settings July 19, 2025 15:02

nberlette added documentation Improvements or additions to documentation enhancement New feature or request labels Jul 19, 2025

refactor(finite): export Finite type from guards/finite

ca6da68

Copilot AI reviewed Jul 19, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

refactor: everything #10

refactor: everything #10

Uh oh!

nberlette commented Jul 19, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Copilot AI Jul 19, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	negative_zero: 0x8000000000000000n,
	negative_zero: 0x80000000000000000000000000000000n,

	* The number of bits used for the exponent in the IEEE-754 half-precision
	* The number of bits used for the exponent in the IEEE-754 single-precision

	* The number of bits used for the mantissa in the IEEE-754 half-precision
	* The number of bits used for the mantissa in the IEEE-754 single-precision

	* The exponent bias used in the IEEE-754 half-precision (or "binary32") format.
	* The exponent bias used in the IEEE-754 single-precision (or "binary32") format.

	* Well-known bit pattern representing `NaN` in the IEEE-754 half-precision (or
	* Well-known bit pattern representing `NaN` in the IEEE-754 single-precision (or

	* IEEE-754 half-precision (or "binary32") format.
	* IEEE-754 single-precision (or "binary32") format.

	* half-precision (or "binary32") format.
	* single-precision (or "binary32") format.

Uh oh!

refactor: everything #10

Are you sure you want to change the base?

refactor: everything #10

Uh oh!

Conversation

nberlette commented Jul 19, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 19, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants