fix N (Issue #103)

[mmatera]

The evaluation is still slow, but this avoids the overflow. There are two things that could be improved:

• to check the size of the number before the conversion, to avoid using an exception.
• the default value of d could be taken from mathics.core.number instead of hardcoding it to 15.

[rocky]

The evaluation is still slow, but this avoids the overflow.

Great!

There are two things that could be improved:

• to check the size of the number before the conversion, to avoid using an exception.

For this, bitlength() is what you want.

• the default value of d could be taken from mathics.core.number instead of hardcoding it to 15.

https://github.com/Mathics3/mathics-core/blob/master/mathics/core/number.py#L28-L35 seems to suggest that the default value would be 17, no? @mmatera where does 15 come into this?

See #107 for a suggestion as to how to use. Feel free to commit over or adapt here otherwise.

[mmatera]

The evaluation is still slow, but this avoids the overflow.

Great!

There are two things that could be improved:

• to check the size of the number before the conversion, to avoid using an exception.

For this, bitlength() is what you want.

• the default value of d could be taken from mathics.core.number instead of hardcoding it to 15.

https://github.com/Mathics3/mathics-core/blob/master/mathics/core/number.py#L28-L35 seems to suggest that the default value would be 17, no? @mmatera where does 15 come into this?

@rocky: I used 15, in order to reproduce what we get using Sympy, but you are right, WMA uses 17.

See #107 for a suggestion as to how to use. Feel free to commit over or adapt here otherwise.

[mmatera]

@rocky, trying to merge your branch into this one, I merge by accident this branch into master. I tried to revert, but it seems it does not work. To avoid more noise, I continue this in #109.

[rocky]

np