Wrong test for log1p ?

[MichaelClerx]

In the mathml evaluation test, we have some extra "functions"

    output local:expm1 units dimensionless
    define local:expm1 = MathML:exp(local:input) - 1 :: dimensionless
    output local:ln units dimensionless
    define local:ln = MathML:ln(local:input)
    output local:log units dimensionless
    define local:log = MathML:log(local:input)
    output local:log1p units dimensionless
    define local:log1p = 1 :: dimensionless + MathML:ln(local:input)

Note that these are not available as functions to the user, e.g. there is no expm1 that users can call

The test for log1p here uses log1p = 1 + log(p), which is different from numpy and C++, who both define a log1p as log1p = log(1 + p) (which makes it the inverse of exp1m)

Should we fix this?

[MichaelClerx]

See also #139 #142

[jonc125]

Probably should, yes!