-module(starfuzz@search).
-compile([no_auto_import, nowarn_unused_vars, nowarn_unused_function, nowarn_nomatch, inline]).
-define(FILEPATH, "src/starfuzz/search.gleam").
-export([new/2, by/2, using/2, with_normalizer/2, with_minimum_score/2, with_limit/2, run/2, strings/2]).
-export_type([match/1, search_builder/1]).

-if(?OTP_RELEASE >= 27).
-define(MODULEDOC(Str), -moduledoc(Str)).
-define(DOC(Str), -doc(Str)).
-else.
-define(MODULEDOC(Str), -compile([])).
-define(DOC(Str), -compile([])).
-endif.

?MODULEDOC(
    " A module providing fuzzy search collections matching capabilities.\n"
    " Collections of custom candidates can be searched in-memory with custom selectors,\n"
    " similarity scoring functions, score thresholds, and limits.\n"
).

-type match(EII) :: {match, EII, float(), integer()}.

-type search_builder(EIJ) :: {search_builder,
        list(EIJ),
        fun((EIJ) -> binary()),
        fun((binary(), binary()) -> float()),
        gleam@option:option(starfuzz@normalize:normalizer()),
        float(),
        gleam@option:option(integer())}.

-file("src/starfuzz/search.gleam", 37).
?DOC(
    " Creates a new SearchBuilder with the list of candidates and a string selector function.\n"
    " Defaults to using Levenshtein similarity, no normalizer, minimum score of 0.0, and no limit.\n"
).
-spec new(list(EIK), fun((EIK) -> binary())) -> search_builder(EIK).
new(Candidates, Selector) ->
    {search_builder,
        Candidates,
        Selector,
        fun starfuzz@similarity:levenshtein/2,
        none,
        +0.0,
        none}.

-file("src/starfuzz/search.gleam", 49).
?DOC(" Configures a custom selector function to retrieve the string value to match from candidates.\n").
-spec by(search_builder(EIN), fun((EIN) -> binary())) -> search_builder(EIN).
by(Builder, Selector) ->
    {search_builder,
        erlang:element(2, Builder),
        Selector,
        erlang:element(4, Builder),
        erlang:element(5, Builder),
        erlang:element(6, Builder),
        erlang:element(7, Builder)}.

-file("src/starfuzz/search.gleam", 54).
?DOC(" Configures the similarity scoring function to use for comparison (e.g. `similarity.jaro_winkler`).\n").
-spec using(search_builder(EIQ), fun((binary(), binary()) -> float())) -> search_builder(EIQ).
using(Builder, Scorer) ->
    {search_builder,
        erlang:element(2, Builder),
        erlang:element(3, Builder),
        Scorer,
        erlang:element(5, Builder),
        erlang:element(6, Builder),
        erlang:element(7, Builder)}.

-file("src/starfuzz/search.gleam", 59).
?DOC(" Configures the normalizer to clean up both query and candidate strings before scoring.\n").
-spec with_normalizer(search_builder(EIT), starfuzz@normalize:normalizer()) -> search_builder(EIT).
with_normalizer(Builder, Normalizer) ->
    {search_builder,
        erlang:element(2, Builder),
        erlang:element(3, Builder),
        erlang:element(4, Builder),
        {some, Normalizer},
        erlang:element(6, Builder),
        erlang:element(7, Builder)}.

-file("src/starfuzz/search.gleam", 64).
?DOC(" Configures the minimum score threshold. Matches below this score are discarded.\n").
-spec with_minimum_score(search_builder(EIW), float()) -> search_builder(EIW).
with_minimum_score(Builder, Min_score) ->
    {search_builder,
        erlang:element(2, Builder),
        erlang:element(3, Builder),
        erlang:element(4, Builder),
        erlang:element(5, Builder),
        Min_score,
        erlang:element(7, Builder)}.

-file("src/starfuzz/search.gleam", 69).
?DOC(" Configures a limit on the number of matches returned.\n").
-spec with_limit(search_builder(EIZ), integer()) -> search_builder(EIZ).
with_limit(Builder, Limit) ->
    {search_builder,
        erlang:element(2, Builder),
        erlang:element(3, Builder),
        erlang:element(4, Builder),
        erlang:element(5, Builder),
        erlang:element(6, Builder),
        {some, Limit}}.

-file("src/starfuzz/search.gleam", 106).
-spec compare_matches(match(EJG), match(EJG)) -> gleam@order:order().
compare_matches(A, B) ->
    case erlang:element(3, A) > erlang:element(3, B) of
        true ->
            lt;

        false ->
            case erlang:element(3, A) < erlang:element(3, B) of
                true ->
                    gt;

                false ->
                    case erlang:element(4, A) < erlang:element(4, B) of
                        true ->
                            lt;

                        false ->
                            gt
                    end
            end
    end.

-file("src/starfuzz/search.gleam", 75).
?DOC(
    " Runs the fuzzy search query against the candidates list.\n"
    " Returns matched candidates sorted descending by score, with original index-based tie-breaking.\n"
).
-spec run(search_builder(EJC), binary()) -> list(match(EJC)).
run(Builder, Query) ->
    Norm_query = case erlang:element(5, Builder) of
        {some, Normalizer} ->
            starfuzz@normalize:apply(Normalizer, Query);

        none ->
            Query
    end,
    Candidates_indexed = gleam@list:index_map(
        erlang:element(2, Builder),
        fun(C, I) -> {C, I} end
    ),
    Matches = gleam@list:fold(
        Candidates_indexed,
        [],
        fun(Acc, Item) ->
            {Candidate, I@1} = Item,
            Raw_val = (erlang:element(3, Builder))(Candidate),
            Val = case erlang:element(5, Builder) of
                {some, Normalizer@1} ->
                    starfuzz@normalize:apply(Normalizer@1, Raw_val);

                none ->
                    Raw_val
            end,
            Score = (erlang:element(4, Builder))(Norm_query, Val),
            case Score >= erlang:element(6, Builder) of
                true ->
                    [{match, Candidate, Score, I@1} | Acc];

                false ->
                    Acc
            end
        end
    ),
    Sorted = gleam@list:sort(Matches, fun compare_matches/2),
    case erlang:element(7, Builder) of
        {some, L} ->
            gleam@list:take(Sorted, L);

        none ->
            Sorted
    end.

-file("src/starfuzz/search.gleam", 124).
?DOC(" Quick direct search over a list of strings using default Levenshtein similarity.\n").
-spec strings(binary(), list(binary())) -> list(match(binary())).
strings(Query, Candidates) ->
    _pipe = new(Candidates, fun(X) -> X end),
    run(_pipe, Query).