[niho]

Parsing a language like Haskell that does not use any parentheses and curly braces is slightly harder than a typical C like language (or Lisp for that matter). Haskell syntax is indentation sensitive, which require the parser to keep track of the indentation when parsing.

[gnufx]

If you're doing combinator parsing, the offside combinator is a few lines. I have one in front of me as part of a demo Haskell-like syntax for Scheme as a command processor. You have to tag the lexemes with position information, but you presumably want that for error-reporting anyhow.

[As others have pointed out, you get the choice of layout in Haskell syntaxes and, as it happens, that Scheme syntax also allowed you to use sexp input.]

[jraph]

The indentation-sensitive part is probably not a big issue though, Python does well :-)

[niho]

Using curly braces or parentheses to signify blocks of code is the trivial and natural way to implement a parser for a programming language. Doing anything else requires a conscious design decision and some effort and extra complexity. The C philosophy is to avoid complexity in the implementation in favor of a slightly more verbose language. The Haskell (ML) philosophy is the exact opposite. These design decisions have been cargo culted by later languages.

[jraph]

My point of view is:

- make it easy to write and to read at the expense of having complexity in the parser / implementation. Programs (or documents) are written and read a lot of times, it's worth optimizing, as long as it does not make the implementation unmaintainable. I'm saying that as a writer of several manually crafted parsers.

- avoid ambiguities in the grammar at all cost. They just suck for every party involved.

As for indentation vs braces to delimit blocks, I practice both and don't really have any preference.

[voldacar]

Copying and pasting python code is annoying (more so because it doesn't have macros or any metaprogramming really) and having to indent things in the python repl just feels dumb. I don't know anyone who thinks python does "well" in this area.

[pmontra]

It does well but I'm not happy when I move code around and have to manually fix the indentation level at destination and check if I inadvertently broke an if or some loop. My editor does it for me with other languages and I have one less source of bugs.

[jraph]

I've noticed editors auto fixing Python indentation while copy-pasting. I think VSCode does it, so it's definitively a solvable problem :-)

(I've disabled this feature in my main editor for a long time though, so I'm used to hit Tab/Shift+Tab as many times as needed when pasting)

[jakear]

VS Code tries to do it, but it’s heuristic based and generally fallible. (Disclaimer, work on vscode, not on python extension)

I much prefer working in languages wherein the series of tokens absolutely defines the semantics (as opposed to Python where the series of tokens + the context define the semantics)

[pmontra]

Yep, but no editor can decide if the code must align with the last line of a loop or be indented out of the loop. I got at least one bug in production because of that, and a test that didn't catch the error.

[gnufx]

For what it's worth, Python doesn't actually use ISWIM's offside rule, as Haskell does; colons are involved.