Well, for starters,

    start env = function
      | [] -> ()
      | line :: lines
        when String.length line > 4 && String.sub line 0 3 = "+++" ->
          header env line;
          modified env 0 lines
      | _ :: lines -> start env lines

is a bit elitist way to write a 2 argument function as a combination of 1 argument function and a lambda function. So let me rewrite:

    start (env) (list) = match list with
      | [] -> ()
      | line :: lines
          when String.length line > 4 && String.sub line 0 3 = "+++" ->
          header env line;
          modified env 0 lines
      | _ :: lines -> start env lines

So the function `start` takes some kind of state `env` as a first argument, and then it takes a list as a second argument, here named `list`.

If the list is empty, [], it does nothing.

If the first list element, which is a string, starts with "+++", it calls two other functions: first header(env,line) where `line` is now the first list element, and then modified(env,0,lines) where `lines` is the rest of the list, not containing the first element.

If the first list element does not start with +++, it recurses to the rest of the list.

So this is not a filter: After first encounter of "+++" it calls those two other functions and stops.

Sounds a bit silly to write something like that by hand, when you could use BatList.drop_while (from batteries) like so:

    let line_is_not_header line =
        String.length line < 4 || String.sub line 0 3 != "+++"
    match (BatList.drop_while line_is_not_header lines) with
    | header_line :: following_lines ->
        header env header_line;
        modified env 0 following_lines
    | _ -> ()

In practice, I find that it's rarely useful to write recursive functions by hand, unless you're willing to limit yourself to the default stdlib.

In python3 something like:

    def start(env, lines):
        while lines:
            line, *lines = lines
            if len(line) > 4 and line.startswith("+++"):
                header(env, line)
                modified(env, 0, lines)

Notes:

    while lines:

is pythonic for

    while not lines == []:

Correction (I forgot the break):

    def start(env, lines):
        while lines:
            line, *lines = lines
            if len(line) > 4 and line.startswith("+++"):
                header(env, line)
                modified(env, 0, lines)
                break

The ocaml standard library is minimal to a fault. It's just enough to implement the compiler itself. For higher level operations like starts_with, you use libraries. The Batteries library for example provides what you want:

    String.starts_with line "+++"

It's a problem. Until people have standardized on which 'better standard library' to use, programs will keep building from these too low level building blocks.

It is a style, not a problem. OCaml had been there a long long time. I've brushed with it a couple of times over the years, and every time the same thing.

You look at some random OCaml code, and it looks like this:

    when String.length line > 4 && String.sub line 0 3 = "+++"

I'm looking at it, and I'm not even sure - is it a bug there? Should it be '>=3' or they've really wanted to say '> 4'. Numbers all other the place? Ten different variants of equal signs?

You can call it a style if that makes a difference. I think there are reasons behind that style, and I think the too minimal standard library is the main reason. The lack of a good package manager until recently may be another contributing reason.

I hesitate to answer your hyperbole about ten variants of equal signs. It's not uncommon for languages to have two notions of equality like ocaml has (or even more): Here's Python for example: http://stackoverflow.com/questions/132988/is-there-a-differe...

To answer your original question of why anyone would prefer to not use startswith: I don't think anyone does.