Rationalize giving error on 1/(x^-2)

[thunderkid]

Running math.rationalize('1/(x^-2)') gives the error Cannot read property '1' of undefined.

However, running math.rationalize('1/(x^-1)') gives the result x as you'd expect.

I'm using mathjs 6.5.0. Haven't tested other versions.

[josdejong]

Hm, that looks like a bug. The stacktrace I get is:

> math.rationalize('1/(x^(-2))').toString()
TypeError: Cannot read property '1' of undefined
    at recPoly (/home/jos/Dropbox/Jos/projects/js/mathjs/lib/function/algebra/rationalize.js:291:24)
    at recPoly (/home/jos/Dropbox/Jos/projects/js/mathjs/lib/function/algebra/rationalize.js:302:13)
    at polynomial (/home/jos/Dropbox/Jos/projects/js/mathjs/lib/function/algebra/rationalize.js:257:5)
    at NodeObjectBoolean (/home/jos/Dropbox/Jos/projects/js/mathjs/lib/function/algebra/rationalize.js:143:21)
    at generic (/home/jos/Dropbox/Jos/projects/js/mathjs/node_modules/typed-function/typed-function.js:1073:27)
    at rationalize (/home/jos/Dropbox/Jos/projects/js/mathjs/node_modules/typed-function/typed-function.js:1092:24)
    at string (/home/jos/Dropbox/Jos/projects/js/mathjs/lib/function/algebra/rationalize.js:120:14)
    at Object.rationalize (/home/jos/Dropbox/Jos/projects/js/mathjs/node_modules/typed-function/typed-function.js:1085:85)

[josdejong]

The cause of this issue is here, in the part if (node.args[1].fn === 'unaryMinus') {...}:

https://github.com/josdejong/mathjs/blob/2a71fe8438b41a8495179953657cdf061b6070c2/src/function/algebra/rationalize.js#L277-L285

When I remove these three lines 278 to 280, no unit tests break and the expression '1/(x^(-2))' gives a normal error Error: There is a non-integer exponent.

@paulobuchsbaum do you still remember what these three lines where supposed to do? If I'm not mistaken you have done some changes there in commit 443d42a7fcea1fb09828e5cf703aa28e2b043005.

[paulobuchsbaum]

Dear Jos de Jong

These 3 lines don't seem to make any sense for me.

Best regards, Paulo Buchsbaum

Em 01/02/2020 13:05:50, "Jos de Jong" [email protected] escreveu:

The cause of this issue is here, in the part if (node.args[1].fn === 'unaryMinus') {...}:

https://github.com/josdejong/mathjs/blob/2a71fe8438b41a8495179953657cdf061b6070c2/src/function/algebra/rationalize.js#L277-L285

When I remove these three lines 278 to 280, no unit tests break and the expression '1/(x^(-2))' gives a normal error Error: There is a non-integer exponent.

@paulobuchsbaum https://github.com/paulobuchsbaum do you still remember what these three lines where supposed to do? If I'm not mistaken you have done some changes there in commit 443d42a https://github.com/josdejong/mathjs/commit/443d42a7fcea1fb09828e5cf703aa28e2b043005.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/josdejong/mathjs/issues/1724?email_source=notifications&email_token=AHI5QJTVVWRNGYTTGBYG32LRAWMV5A5CNFSM4KMIB4IKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKRANRA#issuecomment-581043908, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHI5QJVONV62RMBOPTMFJMLRAWMV5ANCNFSM4KMIB4IA.

[josdejong]

Thanks Paulo. I think the intention of these lines was to maybe turn a unary minus operator with a constant into a constant with a negative value.

I will remove these three lines for now, that will at least give a clear error message.

It would be great if anyone can help out actually implementing support for negative exponents!

[joshhansen]

Just ran into this in 9.1.0, specifically 'There is a non-integer exponent' on math.rationalize('x ^ -2').

I'm looking at trying to make a fix. I have recPoly identifying the unaryMinus negative exponent and throwing Error('There is a negative exponent').

What do we want to happen when this occurs? Should rationalize just return the simplified input expression? It seems to do this on math.rationalize('x ^ -1'), which comes out as 1 / x

Or should rationalize itself propogate the 'There is a negative exponent' error, like happens right now for 'There is a non-integer exponent'? It's notable that this doesn't happen for ('x ^ -1') which simplifies, whereas bigger negative exponents do not.

[josdejong]

Thanks @joshhansen for looking into this.

If possible I think the outcome 1 / x would be best: this is a proper fraction notation which is what rationalize tries to generate. If not possible, a more accurate error message would be nice. I'm not sure at this point whether there is a bug into play here, or that we need to extend the code to be able to handle cases like these.

[gwhitney]

I agree that x^-2 should rationalize to 1/x^2 and 1/(x^-2) to x^2 and yet, almost four years on, still neither works. Hopefully this issue can get some love eventually. ;-)