alternative syntax

[stoeffel]

from discussion in #10

The current implementation is the same as python's doctest.

{-|
    >>> indexOfLineWithEquals
    ...      <| Array.fromList
    ...      <| String.lines "a\nb\nc = d\nf = d\n"
    2
-}

vote 👍

{-|
    > indexOfLineWithEquals
    #      <| Array.fromList
    #      <| String.lines "a\nb\nc = d\nf = d\n"
    2
-}

vote 😄

{-|
    > indexOfLineWithEquals
    ~      <| Array.fromList
    ~      <| String.lines "a\nb\nc = d\nf = d\n"
    2
-}

vote 🎉

{-|
    > indexOfLineWithEquals
    \      <| Array.fromList
    \      <| String.lines "a\nb\nc = d\nf = d\n"
    2
-}

vote ❤️

{-|
    > indexOfLineWithEquals
    |      <| Array.fromList
    |      <| String.lines "a\nb\nc = d\nf = d\n"
    2
-}

vote : 😕 (sorry for the emoji, there aren't many options)

[stoeffel]

I would vote for 🎉 if we decide on going with an alternative solution. I still think that using one character can be confusing and can be mistaken for part of the code, where ... and >>> is more obvious.

[stoeffel]

btw. haskell doctest also uses >>>

[AnthonyJacob]

Why the hell are you trying to invent new syntax when elm-repl has already established one?

> (\a -> a)
| [ 2
| , 3
| , 5
| ]
[2,3,5] : List number

I vote for non of the above but for the elm-repl one!

Or there is another option. Make elm-repl use the new docs syntax you choose instead of its current one!

[despairblue]

@AnthonyJacob good idea to consider the elm repl syntax as well, although the tone of your suggestion can be improved.

I'd also vote for the elm repl syntax.

[stoeffel]

I included repl-style in the poll.
The reason why I don't like the syntax with only one char is, it can easily be confused to be a typo. >>> is more obvious.
The reason why I am against repl-style is, because:

• it implies that elm-doc-test uses the repl to run your tests
• you need \ to start a new line in repl and you don't have to type > or |
• >>> is used in a doctest for other languages (more familiar to newcomers)
• >>> is easier to search

[AnthonyJacob]

I am sorry for that tone, I was going trough some packages-extra made by haskellers and it wasn't pleasant experience :)

you need \ to start a new line in repl

I hope that in the near feature this won't be the case. ( shift+enter is much more convenient )

in repl you don't have to type > or |

Once there will be good IDE support for documentation, you also won't need it. It might feel exactly like the repl actually.

...

All the other things you pointed out can be solved by one bold line in documentation.

On the other hand, if the change would get propagated into repl, I would go with the first one.

[eeue56]

Why the hell are you trying to invent new syntax when elm-repl has already established one?

This tone is not okay. This is an open source project that we have spent time on. We've also asked for people's feedback in multiple places on the syntax.

I vote for non of the above but for the elm-repl one!

The reason why the prompt does not matter is because 1) >>> is not going seem much different from > to beginners 2) doc tests are not completely equivalent to the repl and 3) clear, defining syntax is important in doctest parsing.

Haskell's doctests do not prefix things with Prelude>. In languages like Python, where doctests came from, there is a lot of confusion for beginners when doctests do not behave like the REPL. An example is how strings are handled. Very confusing. Even more confusing considering that multiple languages have different prompts and syntax handling entirely. I think this is a solution to a problem that nobody will have. However, if you can present somebody who's confused by the use of >>> over >, I'm happy to change my opinion.

I am sorry for that tone, I was going trough some packages-extra made by haskellers and it wasn't pleasant experience :)

If you are having a bad day, talk it out with someone else - don't take it out on people who are giving up their free time to work on projects.

I hope that in the near feature this won't be the case. ( shift+enter is much more convenient )
Once there will be good IDE support for documentation, you also won't need it. It might feel exactly like the repl actually.

Wishful thinking. Maybe if we planned this whole library around Elm having reflection then we wouldn't even need to generate elm-tests! Except that's not how things are right now, and to plan on some non-existent tooling changes at some point in the future is not a good way to get things done.

[AnthonyJacob]

Maybe if we planned this whole library around Elm having reflection then we wouldn't even need to generate elm-tests!

So it is easier to write generator instead of making a pull request that would bring reflection into elm?

[eeue56]

Adding reflection is a huge design decision, it's not as simple as just adding a pr. So yes.

[AnthonyJacob]

Is / Was there some discussion on this topic (reflection) ? I wanted to make elm-repl better at exploration of the language / libraries / documentation etc. which would very likely require some amount of reflection ( even if only repl would have permission to use it ).

[stoeffel]

closed in favour of #22