random QQ

[DanGrayson]

Frank Schreyer would like "random QQ" to yield negative numbers about half the time (currently, none are provided) and that the various numbers occur with equal probability (currently, fractions of small height occur more often).

Moreover, when working with polynomial rings over the rationals, it seems senseless to provide any denominators:

i21 : R=QQ[x,y]

o21 = R

o21 : PolynomialRing

i22 : random (10, R)

      3 10   5 9    1 8 2   5 7 3   1 6 4   5 5 5   9 4 6   1 3 7    2 8      9   10 10
o22 = -x   + -x y + -x y  + -x y  + -x y  + -x y  + -x y  + -x y  + x y  + x*y  + --y
      5      7      8       3       7       2       4       7                      9

o22 : R

[mahrud]

I was going to report the same for ZZ, but seems like they are related:

i1 : random(ZZ^3,ZZ^3)

o1 = | 7 3 0 |
     | 6 1 8 |
     | 8 8 4 |

              3        3
o1 : Matrix ZZ  <--- ZZ

[DanGrayson]

One possible problem with changing the behavior of "random" is that tests that use it may fail.

[mikestillman]

This is true, but might be worth the pain.

…

On Sep 19, 2019, at 10:51 AM, Daniel R. Grayson ***@***.***> wrote: One possible problem with changing the behavior of "random" is that tests that use it may fail. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#999?email_source=notifications&email_token=AAAY6R4T3ZEJKNDU3ZF5ZNDQKM4RDA5CNFSM4IXAMRR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD7CXKBY#issuecomment-533034247>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAY6R2YDH6BCE5PGJP6BFLQKM4RDANCNFSM4IXAMRRQ>.

[kschwede]

I've thought about implementing a better random ZZ (and hence QQ) where the number is created out of some distribution on the integers. Is someone else already implementing this or something similar in some package?