Long equations from sympy do not wrap

[gabemorris12]

When converting to a PDF, long equations from cell outputs do not wrap correctly:

It does wrap the equation but it still gets clipped off and extends into the margins. This also happens for big matrices:

A similar issue is seen in #275, but I don't think this is quite the same thing.

Steps to Reproduce
Make a notebook that contains the following cells and run it.

import sympy as sp

t, v0 = sp.symbols('t v0')
x = sp.Function('x')(t)

eq = sp.Eq(2*x.diff(t, 4) + 14*x.diff(t, 3) + sp.Rational(93, 2)*x.diff(t, 2) + 127*x.diff(t) + 148*x, 200*sp.Heaviside(t))
eq

# Producing a really long equation
sol = sp.dsolve(eq, ics={
    x.subs(t, 0): 0,
    x.diff(t, 1).subs(t, 0): v0,
    x.diff(t, 2).subs(t, 0): 0,
    x.diff(t, 3).subs(t, 0): 0
})
sol

# Large expression with shorter terms
terms = [sp.sin(i*t) for i in range(100)]
expr = sum(terms)
expr

# Producing a large matrix expression
x = sp.symbols('x0:10')
x_list = [x[i] for i in range(10)]
x_mat = sp.Matrix(x_list)

y = sp.symbols('y0:10')
y_list = [y[i] for i in range(10)]
y_mat = sp.Matrix(y_list)

A = sp.Matrix(10, 10, lambda i, j: sp.sin(i + j))
expr = sp.MatMul(A, x_mat)
for _ in range(10):
    expr = sp.Add(expr, y_mat)
    expr = sp.Add(expr, x_mat)
expr

Run jupyter nbconvert --to pdf <notebook_name.ipynb>

Then you will see this:

It wraps ok with the long equation that has shorter terms, but in that first cell, there is a really long term that should be brought to the next line. The same is true for the large matrix term in the last output.

[t-makaro]

This is definitely a duplicate of #275. NBconvert simply takes the latex that sympy outputs (which struggles with large subexpressions), and I'm not sure there is a way to fix this given the differences between XeLaTeX and MathJax.

[mgeier]

@gabemorris12 For reproducibility, can you please provide a full example of (copyable) code that produces the problematic equations?

[gabemorris12]

@gabemorris12 For reproducibility, can you please provide a full example of (copyable) code that produces the problematic equations?

I made edits to reproduce this issue.

[mgeier]

I made edits to reproduce this issue.

Thanks!

@t-makaro

I'm not sure there is a way to fix this given the differences between XeLaTeX and MathJax.

There are definitely limits, but I think it can still be improved by what I suggested back then in #275 (comment).

@gabemorris12 You can try it with https://github.com/spatialaudio/nbsphinx, which does a slightly better job at breaking equations (but it also has its limits!):

[gabemorris12]

I made edits to reproduce this issue.

Thanks!

@t-makaro

I'm not sure there is a way to fix this given the differences between XeLaTeX and MathJax.

There are definitely limits, but I think it can still be improved by what I suggested back then in #275 (comment).

@gabemorris12 You can try it with https://github.com/spatialaudio/nbsphinx, which does a slightly better job at breaking equations (but it also has its limits!):

That output is better. I like that it breaks up terms, but I'm wondering how it should handle the case where a single term can be longer than the width of the page. Thanks, and I'll look into using this.

[mgeier]

I'm wondering how it should handle the case where a single term can be longer than the width of the page

As I mentioned in #275 (comment), it looks like \left( (which SymPy generates by default) inhibits line breaks. I guess a similar thing happens with the brackets of matrices, and obviously breaking a long fraction will not work.

There might be ways to change the line breaking behavior, but this is beyond my TeX/LaTeX skills. Maybe you can ask at a place like https://tex.stackexchange.com/? I've gotten very helpful answers there in the past.