Make log also accept a base.

[LeifAndersen]

(log a b) = (/ (log a) (log b)), however it has the potential to run faster.

[mflatt]

The base shouldn't be 1, and I think you want an error from log in that case, instead of an error from /. A base of 0 happens to provoke an errr from log, but you may want a more direct check.

[LeifAndersen]

Ah, good catch. I will add error checking for that. Thanks.

[soegaard]

FWIW for testing fllogb from math/flonum can be used.

[http://docs.racket-lang.org/math/flonum.html?q=log#%28def._%28%28lib._math%2Fflonum..rkt%29._fllogb%29%29](fllogb in math/flonum)

[LeifAndersen]

@soegaard Interesting. Although that function only works with flonums, while the log function only works for fixnums.

Either way, thanks. For the life of me I couldn't find an arbitrary log base function in all of our stdlib.

Does anyone else have any thoughts?

[LeifAndersen]

Woops, I mean, the log function (in racket/base), also works on fixnums.

[LeifAndersen]

Okay, @mflatt and @soegaard , I think its ready to merge. The tests are failing, but that seems to be because they are currently failing in master as well. Thoughts?

[soegaard]

@LeifAndersen @mflatt Perhaps recommend using `fllogb` in case where accuracy is important? https://github.com/racket/math/blob/master/math-lib/math/private/flonum/flonum-log.rkt#L138 Somehow it feels odd to have "almost" duplicated functionality. Should log use the same algorithm as fllogb ? Btw `max-dividing-power` provides similar functionality for bignums. 2017-04-14 3:45 GMT+02:00 Leif Andersen <notifications@github.com>:

…

Okay, @mflatt <https://github.com/mflatt> and @soegaard <https://github.com/soegaard> , I think its ready to merge. The tests are failing, but that seems to be because they are currently failing in master as well. Thoughts? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1667 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAcLxf8dZnd_RDljukudCzub8W3ngFypks5rvs-3gaJpZM4M8DrV> .

-- -- Jens Axel Søgaard

[LeifAndersen]

@soegaard @mflatt

Perhaps recommend using fllogb in case where accuracy is important?

Good call. I'll update the docs to reflect that tonight.

Somehow it feels odd to have "almost" duplicated functionality. Should log use the same algorithm as fllogb ?

I don't care one way or another. To me, it makes sense to have (log a b) defined as (/ (log a) (log b)), but has the potential to be computed faster. But if we want to use the same algorithm as fllogb, I guess that's fine too. It just seems odd to me to have:
(log x) and (log x (exp 1)) return slightly different values because they are using a different algorithm. (The existing log algorithm in the first case and the fllogb algorithm in the second case.)

Thoughts?

One other thing, does anyone have a preference of whether this gets put into 6.9, or should wait until 6.10? I'm okay either way, but if we want 6.9 I should merge this tonight.

[LeifAndersen]

Since it seems like everyone is okay I'm going to go ahead and merge this.