Skip to content

Commit

Permalink
Update README with design docs
Browse files Browse the repository at this point in the history
  • Loading branch information
chouzar committed Oct 24, 2024
1 parent 3987b07 commit cfac675
Showing 1 changed file with 105 additions and 0 deletions.
105 changes: 105 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -22,3 +22,108 @@ Further documentation can be found at <https://hexdocs.pm/lamb>.
gleam run # Run the project
gleam test # Run the tests
```

---

Work in progress notes.

# Table API

Currently viewing the table creating in terms of protection levels:
- `Private`, private table, not named.
- `Protected`, protected table, named.
- `Public`, public table, named.

If we would like more precision than this, maybe having an `options` helper with default options would
help to tinker with tables:

```gleam
pub fn options() -> Options {
https://www.erlang.org/doc/apps/stdlib/ets.html#info/1
todo
}
```

Still need to figure out what is going to be the API for differentiating between `set` and `bag` tables.

# Query API

Matchspecs are composed by a Tuple of arity 3 called a `MatchFunction`:

- A `Head` that contains the shape of the data we want to match to, as well as variable declarations.
- A list of `Condition` expressions that can help filter data through predicates.
- A `Body` that declares the shape and variables we'd like to output from the `MatchFunction`.

```erlang
{Head, [Condition], [Body]}
```

Operating on these 3 pieces may be tackled in different ways.

## Alternative 1

Have an API that composes matchspecs together with a builder pattern:

```gleam
let query =
q.new() // Builds a basic default match function.
|> q.bind(Person) // Modifies the head to match the constructor.
|> q.match(field: 5, "Citizen") // Modifies the head to match the value.
|> q.where(field: 3, op: ">=", than: 18) // Adds a condition expression.
query |> q.select(object()) // Returns the whole object #(index, record).
query |> q.select(index()) // Returns just the index of record.
query |> q.select(#(v(2), v(1))) // Returns the 2nd and 1st variables in a tuple.
query |> q.select(False) // Returns false for each matching record.
query |> q.select(Driver) // Maps the variables to a constructor.
query |> q.select(fn(last, first) { // Alternative mapper to above.
Driver(last, first)
})
```

## Alternative 2

Build a "nice" parser to compose complex queries:

```gleam
let query = "
from Person
select 2, 1
where 'Citizen' = 5
where 3 >= 18
select #(0, 1, 2)
"
```

## Alternative 3

Just build a more straightforward mapper for matchspecs.

```gleam
let head = #(any(), record("Person", v(1), v(2), v(3), "Citizen"))
let condition = #(">=", v(3), 18)
let body = body(record("Driver", v(2), v(1)))
let query_driver_elegibility = #(head, [condition], [body])
lamb.search(table, [query_driver_elegibility])
```

## Notes on types and validation

None of these alternatives currently provide a good way of enforcing types but are meant to fail gracefully if
there are errors. Querying so far is a "dynamic" operation.

An alternative would be to run validations at runtime with the help of a schema but that would
be quite a big lift for my current purposes.

Maybe another way is to provide a validator at init time, but this would exclusively check the validity of the
matchspecs.

# Checkout these ETS APIs

* https://www.erlang.org/doc/apps/stdlib/ets.html#select_delete/2
* https://www.erlang.org/doc/apps/stdlib/ets.html#fun2ms/1
* https://www.erlang.org/doc/apps/stdlib/ets.html#test_ms/2
* https://www.erlang.org/doc/apps/stdlib/ets.html#table/2
* https://www.erlang.org/doc/apps/stdlib/qlc
* https://www.erlang.org/doc/apps/stdlib/erl_parse.html#abstract/1

0 comments on commit cfac675

Please sign in to comment.