belt 3.6.0

use std::{
    collections::HashMap,
    path::{Path, PathBuf},
};

use csv::Reader;

use crate::{
    benchmark::{parser::BenchmarkRun, runner::VerboseData},
    core::{
        error::{BenchmarkErrorKind, Result},
        utils,
    },
};

/// Read both results.csv and *_verbose_metrics.csv and reconstruct the data therein
#[allow(clippy::complexity)]
pub fn read_data(
    data_dir: &Path,
) -> Result<(Vec<BenchmarkRun>, HashMap<String, Vec<VerboseData>>)> {
    let files = utils::find_data_files(data_dir)?;

    let results_csv = files
        .iter()
        .find(|file| file.file_name().and_then(|name| name.to_str()) == Some("results.csv"))
        .ok_or_else(|| BenchmarkErrorKind::DataFileNotFound {
            path: data_dir.join("results.csv"),
        })?;

    let verbose_files: Vec<PathBuf> = files
        .iter()
        .filter(|file| {
            file.file_name()
                .and_then(|name| name.to_str())
                .map(|string| string.ends_with("_verbose_metrics.csv"))
                .unwrap_or(false)
        })
        .cloned()
        .collect();

    let results = read_benchmark_results(results_csv)?;
    let verbose_data = read_verbose_data(&verbose_files)?;

    Ok((results, verbose_data))
}

/// Read results.csv and reconstruct BenchmarkResult vector
fn read_benchmark_results(csv_path: &PathBuf) -> Result<Vec<BenchmarkRun>> {
    let mut reader = Reader::from_path(csv_path)?;
    let mut runs: Vec<BenchmarkRun> = Vec::new();

    for row in reader.records() {
        let record = row?;

        let save_name = record.get(0).unwrap_or("unknown").to_string();
        let index: u32 = record.get(1).unwrap_or("0").parse().unwrap_or(0);

        let execution_time_ms: f64 = record.get(2).unwrap_or("0").parse().unwrap_or(0.0);
        let avg_ms: f64 = record.get(3).unwrap_or("0").parse().unwrap_or(0.0);
        let min_ms: f64 = record.get(4).unwrap_or("0").parse().unwrap_or(0.0);
        let max_ms: f64 = record.get(5).unwrap_or("0").parse().unwrap_or(0.0);
        let effective_ups: f64 = record.get(6).unwrap_or("0").parse().unwrap_or(0.0);
        let base_diff: f64 = record.get(7).unwrap_or("0").parse().unwrap_or(0.0);

        let ticks: u32 = record.get(8).unwrap_or("0").parse().unwrap_or(0);
        let factorio_version = record.get(9).unwrap_or("unknown").to_string();
        let platform = record.get(10).unwrap_or("unknown").to_string();

        runs.push(BenchmarkRun {
            index,
            save_name,
            factorio_version,
            platform,
            execution_time_ms,
            ticks,
            avg_ms,
            min_ms,
            max_ms,
            effective_ups,
            base_diff,
            mimalloc_stats: None,
        });
    }

    // Optional: stable-ish ordering (by save then index), handy for debugging
    runs.sort_by(|a, b| {
        a.save_name
            .cmp(&b.save_name)
            .then_with(|| a.index.cmp(&b.index))
    });

    tracing::debug!("Read results from: {}", csv_path.display());
    Ok(runs)
}

/// Read *_verbose_metrics.csv files and reconstruct VerboseData
fn read_verbose_data(verbose_csv_files: &[PathBuf]) -> Result<HashMap<String, Vec<VerboseData>>> {
    let mut verbose_data_by_save: HashMap<String, Vec<VerboseData>> = HashMap::new();

    for csv_path in verbose_csv_files {
        let file_stem = csv_path
            .file_stem()
            .and_then(|s| s.to_str())
            .unwrap_or("")
            .strip_suffix("_verbose_metrics")
            .unwrap_or("")
            .to_string();

        let mut reader = Reader::from_path(csv_path)?;
        let headers = reader.headers()?.clone();

        let mut runs_data: HashMap<usize, String> = HashMap::new();

        for result in reader.records() {
            let record = result?;
            let run_index: usize = record.get(1).unwrap_or("0").parse().unwrap_or(0);

            let entry = runs_data.entry(run_index).or_insert_with(|| {
                let mut csv_content = String::new();
                let original_headers: Vec<&str> = headers.iter().skip(2).collect();
                csv_content.push_str(&original_headers.join(","));
                csv_content.push('\n');
                csv_content
            });

            let data_values: Vec<&str> = record.iter().skip(2).collect();
            entry.push_str(&format!("{}\n", data_values.join(",")));
        }

        let mut verbose_data: Vec<VerboseData> = runs_data
            .into_iter()
            .map(|(run_index, csv_data)| VerboseData {
                save_name: file_stem.clone(),
                run_index: run_index.try_into().unwrap_or(u32::MAX),
                csv_data,
            })
            .collect();

        verbose_data.sort_by_key(|vd| vd.run_index);
        verbose_data_by_save.insert(file_stem, verbose_data);
    }

    tracing::debug!("Read data from:");
    for file in verbose_csv_files {
        tracing::debug!("  - {}", file.display())
    }

    Ok(verbose_data_by_save)
}