Format scientific notation returns NaN if negative exponential is beyond 7 points

[dtruong1801]

numeral return NaN in version 2.0.6 if format negative exponential is beyond 7 points.

numeral(1e-7).format('0,0.00') version 2.0.4 numeral returns 0.00 version 2.0.6 numeral return NaN

[danawoodman]

Same here 😭

[tiesont]

For helping to troubleshoot: https://jsfiddle.net/86syrqL3/

Happens even with a exponential value produced by numeral itself. Probably worth doing a diff on the two versions to see what changed.

[bushmango]

We are also having this problem!

[JasonRammoray]

The reason is hidden in the numeral.js, row 207:

int = numeral._.toFixed(value, 0, roundingFunction);

, which itself leads us to the implementation of toFixed method, defined in rows 380-406:

toFixed: function(value, maxDecimals, roundingFunction, optionals) {
            var splitValue = value.toString().split('.'),
                minDecimals = maxDecimals - (optionals || 0),
                boundedPrecision,
                optionalsRegExp,
                power,
                output;

            // Use the smallest precision value possible to avoid errors from floating point representation
            if (splitValue.length === 2) {
              boundedPrecision = Math.min(Math.max(splitValue[1].length, minDecimals), maxDecimals);
            } else {
              boundedPrecision = minDecimals;
            }

            power = Math.pow(10, boundedPrecision);

            // Multiply up by precision, round accurately, then divide and use native toFixed():
            output = (roundingFunction(value + 'e+' + boundedPrecision) / power).toFixed(boundedPrecision);

            if (optionals > maxDecimals - boundedPrecision) {
                optionalsRegExp = new RegExp('\\.?0{1,' + (optionals - (maxDecimals - boundedPrecision)) + '}$');
                output = output.replace(optionalsRegExp, '');
            }

            return output;
        }

Suppose we have a value equal to 1e-7 and we want to show it as an integer (this means, that we set maxDecimals to zero). If we don't pass any optionals (and we don't pass any optionals at all in row 207), then minDecimals will be equal to zero, which means, that boundedPrecision will be also equal to zero, because splitValue is an array consiting of only one value (string '1e-7' in this case). Then we calculate the value for power variable, which is equal to 1. And then we hit the problem itself. If we are using 1e-6, then this expression would give us '0':

output = (roundingFunction(value + 'e+' + boundedPrecision) / power).toFixed(boundedPrecision);

Which is in our case:

output = (Math.round(1e-6 + 'e+' + 0) / 1).toFixed(0) // gives '0'

But once we use 1e-7 we might be suprpised with the outcome, because

output = (Math.round(1e-7 + 'e+' + 0) / 1).toFixed(0) // gives 'NaN'

Actually it's pretty logic:

1e-7 + 'e+' + 0 === '1e-7e+0' // the string on the right hand side passed to Math.round results in NaN

[JasonRammoray]

I see, that there is a fix already for that issue, but build on Travis has been failed. That is why this pull-request is still opened.

@RaviDasari, any progress on that?

[RaviDasari]

@JasonRammoray , Yes. I fixed an extra comma in test case which is failing the build and it was green again. But I didn't see any updates from the authors. I am still waiting and I see so many PRs (from a very long time) in queue waiting to be merged by author.

[JasonRammoray]

@RaviDasari , perhaps, it make sense to indicate authors who shall review your request. I see, that "Reviewers" column is empty. I recommend to add at least @adamwdraper as a primary reviewer. From my side I'll also put a comment in your pr with a request to review it as soon as possible.

[RaviDasari]

@JasonRammoray , I am not able to add reviewers because of author settings! I see many PRs(all that I checked just now) don't have reviewers assigned. Perhaps only @adamwdraper can add reviewers to delegate work.

Nevertheless, @adamwdraper should be notified because he is the author and we mentioned him many times.

[JasonRammoray]

@RaviDasari, ok got it. Let's wait for a while.

[michaelvolz]

Any news?

[RaviDasari]

Nope.

On Thu, Sep 21, 2017 at 7:03 AM Michael A. Volz (Flynn) < [email protected]> wrote:

Any news?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/adamwdraper/Numeral-js/issues/512#issuecomment-331124531, or mute the thread https://github.com/notifications/unsubscribe-auth/AF2xypOVN4dKkO7GhrxEvz8kHSXN3vSFks5skkKGgaJpZM4NlgT_ .

-- Thanks & Regards Ravi Babu Dasari

[turnerniles]

I'm not sure how we could get in touch with @adamwdraper to add more collaborators to the project but I tweeted him...

[RaviDasari]

Hmm. On Mon, Dec 4, 2017 at 12:52 AM Niles Turner [email protected] wrote:

I'm not sure how we could get in touch with @adamwdraper https://github.com/adamwdraper to add more collaborators to the project but I tweeted him...

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/adamwdraper/Numeral-js/issues/512#issuecomment-348866723, or mute the thread https://github.com/notifications/unsubscribe-auth/AF2xyhUhinikCrDCXtg8DB8s1G-VnJ8Uks5s84h_gaJpZM4NlgT_ .

[artem-karnaukh]

Any news on this?

[turnerniles]

No news... It seems like either find another way to get in contact with @adamwdraper or find someone with who has the capacity to fork the repository and start a new repo / package.

[tab00]

Has @adamwdraper died, or been ignoring us?

What's the best thing to do now? I use the NPM package.

[JasonRammoray]

@tab00, I wouldn't make such guesses with respect to the author since anything could happen and we don't know that for sure. I am not saying it is good to ignore all the issues this repository has, but lets focus on the productive ways of solving the issue. There are several alternative solutions:

1. roll back to the package version, which suits your project needs and makes all unit tests green
2. [applicable if point 1) failed or you want to control the code by yourself] fork the repo, make changes, cover them with unit tests and fix all other tests in case if they will be broken. Then publish new npm package with similar name (you might also want to make it scoped)
3. try to contact with the author @adamwdraper by any other means (phone, emails, messages in social networks)

[pqvst]

This issue has been bugging me for too long now. numbro seems to handle this fine, so I will probably switch to that instead.

[anton6]

Why is this issue still open? I am switching to "2.0.4" for now, but is this repo still active?

[SourceCipher]

This library should be marked as no longer active/maintained since the last release was in 2017 and all the bugs are ignored. I would recommend using another library which is still maintained.

[eidonjoe]

This library should be marked as no longer active/maintained since the last release was in 2017 and all the bugs are ignored. I would recommend using another library which is still maintained.

@SourceCipher Do you have any recommend?

[SourceCipher]

This library should be marked as no longer active/maintained since the last release was in 2017 and all the bugs are ignored. I would recommend using another library which is still maintained.

@SourceCipher Do you have any recommend?

We have replaced with the https://numbrojs.com library most of our code base

[eradman]

@SourceCipher Do you have any recommend?

We have replaced with the https://numbrojs.com library most of our code base

Numbro also has a fatal flaw, namely rounding to integers does not work https://github.com/BenjaminVanRyseghem/numbro/issues/745