Algorithmic decorator (implement output for _repr_latex_)

[ZibingZhang]

Overview

Adds a IPythonAlgorithmicCodegen class which converts algorithms into $\LaTeX$ for IPython.

References

continues work on #57

Blocked by

None

[ZibingZhang]

A point of interest is that the result of get_latex(style=EXPRESSION) doesn't return valid $\LaTeX$, while get_latex(style=ALGORITHMIC) does. Feels weird to have this discrepancy, but I'm not sure of a good way to resolve it.

[odashi]

General comments before providing detailed review:

• Could you split the pull request into at least 2 separate PRs: introducing IPython-specific codegen and revising frontends?
• IPythonAlgorithmicCodegen makes sense more than `AlgorithmicJupyterCodegen. _repr_latex_ is the feature of IPython and Jupyter (and other REPL runtimes) borrows this interface.
• \displaystyle can be avoided at this point.
• \hspace{} with the em unit is better to follow the size of the characters.

[odashi]

Dummy, please re-request review after all changes in frontend were split into other PR.

[ZibingZhang]

Dummy, please re-request review after all changes in frontend were split into other PR.

I'm not ready for this PR to be reviewed, but #164 is ready for review.

[odashi]

• Use \textsc for small capital letters.
• Wrap all statements by \begin{array}{l} and \end{array}. newlines are unexpected in the bare math mode and it brings unexpected consequence (e.g., $a\\b$ and $$a\\b$$ may generate different results in terms of the number of lines)

[ZibingZhang]

• Use \textsc for small capital letters.

Tried it in the Google collab notebook and \textsc{fact} dosn't seem to be working

Maybe \texttt?

[odashi]

\textsc{fact} dosn't seem to be working

Okay, so I guess \mathrm{...} is eventually better, and I don't agree with any options with manual uppercasing.

[odashi]

\texttt

No, we don't have any necessity to use typewriter font.

[ZibingZhang]

I don't agree with any options with manual uppercasing.

Sounds good, I can remove that. Only had it because ALGORTIHMIC codegen produces \function{fun_name} which auto uppercases.

[odashi]

Restore the original expression, and don't use "+" concatenation as well as possible. f-strings or join() are basically the most fastest option to concatenate strings and "+" is the worst choice.

Ditto for all other places.

def f():
  for i in range(1_000_000):
    x = f"The answer is {i}."
  return x

def g():
  for i in range(1_000_000):
    x = "The answer is " + str(i) + "."
  return x

%time f()
%time g()

CPU times: user 191 ms, sys: 0 ns, total: 191 ms
Wall time: 192 ms
CPU times: user 327 ms, sys: 0 ns, total: 327 ms
Wall time: 327 ms

[ZibingZhang]

Good to know, I think I've converted it all back. When codegen though lots of places can't use f strings because of \ not allowed :(