//! Contains the database implementation of the plant layer.

use std::cmp::max;

use chrono::NaiveDate;
use diesel::{
    debug_query,
    pg::Pg,
    sql_types::{Array, Date, Float, Integer, Uuid as SqlUuid},
    CombineDsl, ExpressionMethods, QueryDsl, QueryResult, QueryableByName,
};
use diesel_async::{AsyncPgConnection, RunQueryDsl};
use log::{debug, trace};
use uuid::Uuid;

use crate::{
    model::{
        dto::{RelationDto, RelationSearchParameters, RelationsDto},
        r#enum::spatial_relation_type::SpatialRelationType,
    },
    schema::spatial_relations,
};

/// A bounding box around the maps geometry.
#[derive(Debug, Clone, QueryableByName)]
struct BoundingBox {
    /// The lowest x value in the geometry.
    #[diesel(sql_type = Integer)]
    x_min: i32,
    /// The lowest y value in the geometry.
    #[diesel(sql_type = Integer)]
    y_min: i32,
    /// The highest x value in the geometry.
    #[diesel(sql_type = Integer)]
    x_max: i32,
    /// The highest y value in the geometry.
    #[diesel(sql_type = Integer)]
    y_max: i32,
}

/// Stores the score of a x,y coordinate on the heatmap.
#[derive(Debug, Clone, QueryableByName)]
struct HeatMapElement {
    /// The score on the heatmap.
    #[diesel(sql_type = Float)]
    preference: f32,
    /// The alpha on the heatmap.
    #[diesel(sql_type = Float)]
    relevance: f32,
    /// The x values of the score
    #[diesel(sql_type = Integer)]
    x: i32,
    /// The y values of the score.
    #[diesel(sql_type = Integer)]
    y: i32,
}

/// Generates a heatmap signaling ideal locations for planting the plant.
///
/// # Errors
/// * If no map with id `map_id` exists.
/// * If no layer with id `layer_id` exists, if the layer is not a plant layer or if the layer is not part of the map.
/// * If no plant with id `plant_id` exists.
#[allow(
    clippy::cast_sign_loss,             // ok, because we will never reach number high enough where this will matter
    clippy::indexing_slicing,           // ok, because we know the size of the matrix using the maps bounding box
    clippy::cast_possible_truncation,   // ok, because ceil prevents invalid truncation
)]
pub async fn heatmap(
    map_id: i32,
    plant_layer_id: Uuid,
    shade_layer_id: Uuid,
    plant_id: i32,
    date: NaiveDate,
    conn: &mut AsyncPgConnection,
) -> QueryResult<Vec<Vec<(f32, f32)>>> {
    // Fetch the bounding box x and y values of the maps coordinates
    let bounding_box_query =
        diesel::sql_query("SELECT * FROM calculate_bbox($1)").bind::<Integer, _>(map_id);
    debug!("{}", debug_query::<Pg, _>(&bounding_box_query));
    let bounding_box = bounding_box_query.get_result::<BoundingBox>(conn).await?;

    let granularity = calculate_granularity(&bounding_box);

    // Fetch the heatmap
    let query =
        diesel::sql_query("SELECT * FROM calculate_heatmap($1, $2, $3, $4, $5, $6, $7, $8, $9)")
            .bind::<Integer, _>(map_id)
            .bind::<Array<SqlUuid>, _>(vec![plant_layer_id, shade_layer_id])
            .bind::<Integer, _>(plant_id)
            .bind::<Date, _>(date)
            .bind::<Integer, _>(granularity)
            .bind::<Integer, _>(bounding_box.x_min)
            .bind::<Integer, _>(bounding_box.y_min)
            .bind::<Integer, _>(bounding_box.x_max)
            .bind::<Integer, _>(bounding_box.y_max);
    debug!("{}", debug_query::<Pg, _>(&query));
    let result = query.load::<HeatMapElement>(conn).await?;

    // Convert the result to a matrix.
    // Matrix will be from 0..0 to ((x_max - x_min) / granularity)..((y_max - y_min) / granularity).
    let num_cols =
        (f64::from(bounding_box.x_max - bounding_box.x_min) / f64::from(granularity)).floor();
    let num_rows =
        (f64::from(bounding_box.y_max - bounding_box.y_min) / f64::from(granularity)).floor();
    let mut heatmap = vec![vec![(0.0, 0.0); num_cols as usize]; num_rows as usize];
    for HeatMapElement {
        preference,
        relevance,
        x,
        y,
    } in result
    {
        heatmap[y as usize][x as usize] = (preference, relevance);
    }

    trace!("{heatmap:#?}");
    Ok(heatmap)
}

/// The number of values the resulting heatmap matrix should have.
const NUMBER_OF_SQUARES: f64 = 10000.0;

/// Calculate granularity so the number of scores calculated stays constant independent of map size.
fn calculate_granularity(bounding_box: &BoundingBox) -> i32 {
    let width = bounding_box.x_max - bounding_box.x_min;
    let height = bounding_box.y_max - bounding_box.y_min;

    // Mathematical reformulation:
    // width * height = number_of_squares * granularity^2
    // granularity = sqrt((width * height) / number_of_squares)
    #[allow(clippy::cast_possible_truncation)] // ok, because we don't care about exact values
    let granularity = (f64::from(width * height) / NUMBER_OF_SQUARES).sqrt() as i32;

    max(1, granularity)
}

/// Get all spatial relations of a certain plant.
///
/// # Errors
/// * If the SQL query fails.
pub async fn find_relations(
    search_query: RelationSearchParameters,
    conn: &mut AsyncPgConnection,
) -> QueryResult<RelationsDto> {
    let query = spatial_relations::table
        .select((spatial_relations::plant2, spatial_relations::relation))
        .filter(spatial_relations::plant1.eq(&search_query.plant_id))
        .union(
            spatial_relations::table
                .select((spatial_relations::plant1, spatial_relations::relation))
                .filter(spatial_relations::plant2.eq(&search_query.plant_id)),
        );
    debug!("{}", debug_query::<Pg, _>(&query));
    let relations = query
        .load::<(i32, SpatialRelationType)>(conn)
        .await?
        .into_iter()
        .map(|(id, relation)| RelationDto { id, relation })
        .collect();
    Ok(RelationsDto {
        id: search_query.plant_id,
        relations,
    })
}