Feature GH-12143: Extend the maximum precision round can handle by one digit

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?

[SakiTakamachi]

Please check here for more details.
#12143 (comment)

In the existing implementation, for example, the following code does not perform rounding:

<?php
var_dump(round(0.12345678901234565, 16, PHP_ROUND_HALF_UP));
// float(0.12345678901234565)
// I want it to be like this
// float(0.1234567890123457)

In this example, the maximum value that can be specified for round() argument $precision is 15.

This pull request changes round() to accept one more fine digit. (16 for the above example)

---

The test will probably not succeed unless it incorporates the changes in #12220.
Also, this PR also includes changes in #12162.

Therefore, you may not be able to accurately check this PR until after these two changes have been merged.

[TimWolla]

@SakiTakamachi Can you rebase this, now that #12220 is merged?

[SakiTakamachi]

@TimWolla
done.

[SakiTakamachi]

The digit adjustment is not going well, I am currently fixing it.

[SakiTakamachi]

This is a malfunction caused by expanding the accepted digits, so there is no problem with the digit adjustment fix PR itself.

[Girgias]

One thing that would make it easier to see what changes in these sorts of PRs is to use the rules LLVM has for PRs.

First commit the test with the current behaviour, then do a follow-up commit with the C changes and how it impacts the test.

[SakiTakamachi]

Thanks, I'll try that

[SakiTakamachi]

@Girgias
First, I created a PR just to add a test to the existing implementation.
#13399

[Girgias]

LGTM, could you add an UPGRADING note when merging?