Help writing a minimal macro

Learning
1

Consider the following code. I define a type once, then I repeat myself 3 times

  • create table
  • first part of insert values
  • second part of insert values

This then happens for EVERY struct/table I define. This is quite error prone, as I’m looking at tens of structs/tables.

In theory, I believe ppx can help me reduce this type of repetition. However, I’m new to ppx. Can someone help with a minimal sample macro? I’m mainly looking for a minimal “skeleton” I can then hack on.

module Row__Browser__Self_Event = struct
  type t = {
    self_id: Fixed_String_16.t;
    self_count: int;
    self_time: Time_Milli.t;

    tag_0: Fixed_String_16.t;
    tag_1: Fixed_String_16.t;
    tag_2: Fixed_String_16.t;

    data: string;
  };;

  let drop_tbl = {sql| DROP TABLE IF EXISTS main_logs.browser_self |sql};;

  let create_tbl = {sql|
  CREATE TABLE IF NOT EXISTS main_logs.browser_self
  (
      self_id          FixedString(16),
      self_count       UInt32,
      self_time_ymd    UInt32,
      self_time_hmsm   UInt32,

      tag_0            FixedString(16),
      tag_1            FixedString(16),
      tag_2            FixedString(16),

      data             String,
  ) ENGINE = MergeTree()
  ORDER BY ()
  |sql};;

  let insert_into = {sql|
  INSERT INTO main_logs.browser_self ( self_id, self_count, self_time_ymd, self_time_hmsm, tag_0, tag_1, tag_2, data) VALUES
  |sql};;


  let insert_values (v: t) (module Q: Single_Quote_Util) = String.concat " " [ 
    "(";
    Q.quote_string v.self_id.raw;
    Q.quote_int v.self_count;
    Q.quote_int v.self_time.year_month_day;
    Q.quote_int v.self_time.hour_minute_second_milli;

    Q.quote_string v.tag_0.raw;
    Q.quote_string v.tag_1.raw;
    Q.quote_string v.tag_2.raw;

    Q.quote_string v.data;
    ")"
  ];;


end
2

EDIT: Better minimal example:

module Row__Browser__Min = struct
  type t = {
    self_count: int;
    data: string;
  };;

  let drop_tbl = {sql| DROP TABLE IF EXISTS main_logs.browser_self |sql};;

  let create_tbl = {sql|
  CREATE TABLE IF NOT EXISTS main_logs.browser_self
  (
      self_count       UInt32,
      data             String,
  ) ENGINE = MergeTree()
  ORDER BY ()
  |sql};;

  let insert_into = {sql|
  INSERT INTO main_logs.browser_self ( self_count, data) VALUES
  |sql};;

  let insert_values (v: t) (module Q: Single_Quote_Util) = String.concat " " [ 
    "(";
    Q.quote_int v.self_count;
    Q.quote_string v.data;
    ")"
  ];;
end

I think this strips out all non-primitive types, yet still captures the “core” of the problem that there is one definition + 3 repetitions, and the 3 repetitions can be replaced via a macro.

3

Is there a readable impl of @@deriving fields from Records - Real World OCaml ? Understanding that macro should suffice for everything I need.

4

The ppxlib repo has two examples, one of which is a deriver, which derives accessor functions from a record type: ppxlib/examples/simple-deriver at main · ocaml-ppx/ppxlib · GitHub

If you are new to PPX, you might be interested in Preprocessors and PPXs · OCaml Tutorials (for some general introduction to PPXs) and Documentation · ppxlib 0.29.1 · OCaml Packages (for some material about writing PPXs).

2 Likes
5

@panglesd : Thanks for the instructive example. I typed up the code, and after fixing many typos I introduced, the code compiled.

For debugging purposes, how do I “printf” the output of the macro?

I.e. I want the output after macro expansion, but before it starts rewriting into whatever IR it uses.

I’m using dune for all my projects.

Thanks!

6

If you mean you want to see the code that the ppx generates (ie after expansion), you can use

$ ocamlc -stop-after parsing -dsource <file.pp.ml> > blah.generated.ml

That file.pp.ml will be in the _build dir.

Edit: this article from Tarides is a nice ppx intro if you’re looking for more reading materials.

1 Like
8

This is insane. I did not know you could specify SHELL SCRIPTS as “preprocessors” in dune. Truly fascinating.

1 Like