How could we improve the syntax?

[paolobrasolin]

@tetrapharmakon suggests that repetitions quickly make code less readable

\mor A ["h",below]:[bend right,->] B;
\mor * ["k", above]:[bend left,->] *;
\mor * ["\sigma\Uparrow",black,mid]:[->,white] *;

and suggests the haskellism

\mor A ["h",below]:[bend right,->] <>
       ["k", above]:[bend left,->] <>
       ["\sigma\Uparrow",black,mid]:[->,white] B;

I feel a better (=clearer, less rigid) working alternative would be

\mor A ["h",below]:[bend right,->] B;
\mor A ["k", above]:[bend left,->] B;
\mor A ["\sigma\Uparrow",black,mid]:[->,white] B;

and an hypothetical improvement (more TikZ-like than the haskellism) could be

\mor A ["h",below]:[bend right,->]
       ["k", above]:[bend left,->]
       ["\sigma\Uparrow",black,mid]:[->,white] B;

Thus some questions are raised:

• how can the syntax be improved staying in line with the TikZ way?
• how can the syntax be improved (redesigned?) to make compositionality apparent?
• are these two directions in contrast?

[tetrapharmakon]

@tetrapharmakon suggests that repetitions quickly make code less readable

\mor A ["h",below]:[bend right,->] B;
\mor * ["k", above]:[bend left,->] *;
\mor * ["\sigma\Uparrow",black,mid]:[->,white] *;

The white arrow is just a hack to draw a 2-cell.

The problem here is just that at the moment there is no way to draw parallel arrows avoiding repetition. Consider the diagram

\obj{A & B};
\mor A -> B;
\mor A [shove=.5em,->] B;
\mor A [shove=-.5em,->] B;

it would be way better (=more readable) to write

\obj{A & B};
\mor A              ->  <>
       [shove=.5em, ->] <>
       [shove=-.5em,->]    B;

1. The code is more rigid, but more readable
2. The whole point of koDi (if I remember) was to have easy to fix/visualize code for diagrams. This style of typesetting is very near to the actual diagram,

   A -->
   -->
   --> B

   is very similar to

   -->
   A --> B
   -->

[paolobrasolin]

@tetrapharmakon suggests that repetitions quickly make code less readable

\mor A ["h",below]:[bend right,->] B;
\mor * ["k", above]:[bend left,->] *;
\mor * ["\sigma\Uparrow",black,mid]:[->,white] *;

The white arrow is just a hack to draw a 2-cell.

This is a less hackish 2-cell, perhaps:

\tikzset{2>/.style={-Implies, shorten >=2pt, shorten <=3pt, double equal sign distance}}

\begin{kodi}
  \obj { A & B \\ };
  \mor[below]:[bend right] A h:-> B;
  \mor[above]:[bend left]  A k:-> B;
  \mor h \sigma:2> k;
\end{kodi}

---

2. The whole point of koDi (if I remember) was to have easy to fix/visualize code for diagrams. This style of typesetting is very near to the actual diagram,

Yup. I'm not sure about the <>, though. Do you have any specific reason for desiring it?

Currently,

• whitespace in an arrow label requires wrapping it in ""
• whitespace in an option (for both arrows and labels) requires wrapping it in []
• whitespace in an object reference requires wrapping it in ()

In fact, whitespace is relegated to the role of separator between an object reference (e.g. (A)) and a labels:arrow block (e.g. ["f"]["g",below]:[->,red]). The latter admits many (or zero) option lists for labels on the left and a single (mandatory) option list on the right for the arrow.

This means that there would be no ambiguity in expressions like

\mor A -> -> -> B;
\mor A -> ->,red [->,light blue] B;
\mor A ["f"]["g", below]:[bend right, ->] ["h"]["i", above]:[bend left, ->] B;
\mor A ["h", below]:[bend right, ->]
       ["k", above]:[bend left, ->] B;

Ignoring the fact that parsing is harder without it, does the <> add something I'm missing?

[paolobrasolin]

Nevermind. Not using any kind of separator is ambiguous:

\mor S [...][...]:[->] T;
% The first block could be both a label and an arrow:
\mor S ["f"]["g"]:[->] T;
\mor S [->] ["g"]:[->] T;

I wouldn't mind a simple comma to concatenate arrows:

\mor S f:-> T g:->, h:-> U i:->, j:->, k:-> V;

That would entail losing the chance to use non delimited arrow option lists:

\mor S f:->,red T;
would be interpreted as
\mor S ["f"]:[->] []:[red] T;
instead of
\mor S ["f"]:[->,red] T;

That might be for the best, though: I think it

• is not a great ergonomics loss
• helps a lot with parsing parallel arrows
• forces to use delimiters for arrow option lists with more than an item -- or a single user defined key (both feel like better code to me)

The initial use case would look like

\mor A ["h",below]:[bend right,->],
       ["k", above]:[bend left,->],
       ["\sigma\Uparrow",black,mid]:[->,white] B;

What do you think, user zero? @tetrapharmakon