Simplifications sometimes fail

[jack-t]

Description

Simple simplifications sometimes fail. Here are some examples:

Correct:

In[21]:= 4 Sqrt[12] + 5 Sqrt[5] + 105 Sqrt[12]
Out[21]= 5 Sqrt[5] + 218 Sqrt[3]

Should be 50 Sqrt[5] + 218 Sqrt[3]:

In[22]:= 4 Sqrt[12] + 50 Sqrt[5] + 105 Sqrt[12]
Out[22]= 8 Sqrt[3] + 50 Sqrt[5] + 210 Sqrt[3]

How to Reproduce

I just opened a Mathics REPL and typed commands in. I observed the same behavior when I wrote the same commands in a script.

Expected behavior

Choosing relatively prime x and y, I wrote expressions of the form a Sqrt[x] + b Sqrt[y] + c Sqrt[x], varying the coefficients. Depending on some relationship between the radicands, a sufficiently large b (see 5 and 50 in the example) seemed to produce this behavior for lots of values of x and y.

Your Environment

Mathics 6.0.4
on CPython 3.9.6 (default, Jul 26 2021, 11:40:16) 
using SymPy 1.12, mpmath 1.3.0, numpy 1.24.3, cython Not installed

[rocky]

I just had a chance to look at this. (I will be pretty busy until the beginning of May).

First of all, the result is not wrong. The result just isn't simplified as you expect. I am not seeing anywhere in WMA docs that say that results have to be automatically simplified.

If you surround the result in an explicit Symplify[] call, you do get the results you expect. So that is a workaround.

@mmatera your thoughts?