Hacker News Reader

Comment by breck | Hacker News Reader

> Which is about 99.9% of the population.

Agreed! For decades some people have said parens in S-Expressions are useless.

But a key thing we're pushing, that's never been pushed before, is that the parens are worse than useless--they are harmful! They prevent easier program concatenation and ad hoc parser writing. And they make program synthesis, visual programming, and code analysis ~10x harder, for next to no gain.

I'm hoping we'll start to turn the tide and bring the 99.9% of the population into the Tree Notation world, and those LISPers that love their parens can remain happily in their (world).

View on HN

> But a key thing we're pushing, that's never been pushed before, is that the parens are worse than useless--they are harmful! They prevent easier program concatenation and ad hoc parser writing. And they make program synthesis, visual programming, and code analysis ~10x harder, for next to no gain.

Almost everything you've said above is not just untrue, but the opposite of the truth. As such, your statements require supporting evidence. Please cite some.

And who, by the way, is "we"?

There is a lot of evidence in the GitHub. See JTree library and the ohayo data vis tool.

There is also a big open source release coming soon—the largest database of programming languages and features in the world by a factor of at least 10x, more like 100x. Over 10k languages, over 1k columns.

We is the Tree Notation Lab. Will be announcing what that is and the funding for it later this month.

> They prevent easier program concatenation and ad hoc parser writing.

This assertion is patently false. Parens make it quite trivial to parse and render the underlying tree structuee rather obvious. Concatenating programs is just a matter of splicing a subtree into the target tree.

> This assertion is patently false.

Here's a proof.

    - In S-Expressions with parens, there are infinite ways to write any given abstract syntax tree.
    - In Tree Notation, there is only 1 way to write any given abstract syntax tree.
      If you change one letter, you get a different tree.
    - Say I wrote programs in a Tree Language with a root node called "dumpToConsole" and I want to
    rename all such nodes to "print". I can write a type-safe ad hoc parser that does
    a simple search replace "/\ndumpToConsole/print/",
    using Tree Notation. If it were in S-Expressions, I'd have to use a full parser.

Reply on news.ycombinator.com