Add tests for "Intl NumberFormat v3" proposal #3307
Conversation
This patch is intended to cover only one aspect of the proposal for ECMA402: the "interpret strings as decimals" feature.
| ['Infinity', 'Infinity'], | ||
| ['-Infinity', '-Infinity'] |
There was a problem hiding this comment.
This doesn't test much; the test is
toNumberResults.forEach(pair => {
const [value, result] = pair;
assert.sameValue(nf.format(value), nf.format(result));
});
so you're only testing that nf.format('Infinity') is the same value as nf.format('Infinity'), which is an identity.
There was a problem hiding this comment.
Ah, that's right. I've changed the second element in each case to be the Number representation of Infinity.
| // The value 100,000 should only be interpreted as infinity if the input is the | ||
| // string "Infinity". | ||
| assert.sameValue(nf.format('100000'), '100,000'); | ||
| // The value -100,000 should only be interpreted as negative infinity if the | ||
| // input is the string "-Infinity". | ||
| assert.sameValue(nf.format('-100000'), '-100,000'); |
There was a problem hiding this comment.
I don't understand what invariant these tests are verifying. Is the number 100,000 significant in the spec somewhere?
There was a problem hiding this comment.
This is my attempt to verify the following conditions in ToIntlMathematicalValue:
- If mv is 1010000 and str contains Infinity, return +∞.
- If mv is -1010000 and str contains Infinity, return -∞.
The intent of these assertions is to ensure that implementations reference not only the mathematical value of the input but also the original string representation. I believe these assertions would fail if an engine incorrectly implemented somethng like this:
- If mv is 1010000, return +∞.
- If mv is -1010000, return -∞.
Does that seem right to you?
| // input is the string "-Infinity". | ||
| assert.sameValue(nf.format('-100000'), '-100,000'); | ||
|
|
||
| assert.sameValue(nf.format('1.0000000000000001'), '1.0000000000000001'); |
There was a problem hiding this comment.
is the result '1.0000000000000001' here?
I think it should be '1' becaues the rounding w/ the default values of fractionDigits and signification digits.
| assert.sameValue(nf.format('-100000'), '-100,000'); | ||
|
|
||
| assert.sameValue(nf.format('1.0000000000000001'), '1.0000000000000001'); | ||
| assert.sameValue(nf.format('-1.0000000000000001'), '1.0000000000000001'); |
There was a problem hiding this comment.
This is surely wrong, a negative number in the string should not be formatted as a posotive number here and without rounding.
|
To close the loop: @FrankYFTang's comments were addressed in #3335 |
This patch is intended to cover only one aspect of the proposal for
ECMA402: the "interpret strings as decimals" feature.
The proposal text for this feature is still being refined; these tests reflect my understanding of the intention based on this discussion in the proposal's repository.
@sffc would you mind taking a look?