use std::{
    collections::{hash_map::Entry, HashMap},
    ffi::OsStr,
    fmt,
    fmt::{Display, Formatter},
    io,
    path::Path,
    time::Instant,
};

use anyhow::{anyhow, bail};
use git_repository as git;
use git_repository::{actor, bstr::BString, interrupt, objs, prelude::*, progress, refs::file::ReferenceExt, Progress};
use itertools::Itertools;
use rayon::prelude::*;

/// Additional configuration for the hours estimation functionality.
pub struct Context<W> {
    /// Show personally identifiable information before the summary. Includes names and email addresses.
    pub show_pii: bool,
    /// Omit unifying identities by name and email which can lead to the same author appear multiple times
    /// due to using different names or email addresses.
    pub omit_unify_identities: bool,
    /// Where to write our output to
    pub out: W,
}

/// Estimate the hours it takes to produce the content of the repository in `_working_dir_`, with `_refname_` for
/// the start of the commit graph traversal.
///
/// * _working_dir_ - The directory containing a '.git/' folder.
/// * _refname_ - The name of the ref like 'main' or 'master' at which to start iterating the commit graph.
/// * _progress_ - A way to provide progress and performance information
pub fn estimate<W, P>(
    working_dir: &Path,
    refname: &OsStr,
    mut progress: P,
    Context {
        show_pii,
        omit_unify_identities,
        mut out,
    }: Context<W>,
) -> anyhow::Result<()>
where
    W: io::Write,
    P: Progress,
{
    let repo = git::discover(working_dir)?.apply_environment();
    let commit_id = repo
        .refs
        .find(refname.to_string_lossy().as_ref())?
        .peel_to_id_in_place(&repo.refs, |oid, buf| {
            repo.objects
                .try_find(oid, buf)
                .map(|obj| obj.map(|obj| (obj.kind, obj.data)))
        })?
        .to_owned();

    let (all_commits, is_shallow) = {
        let start = Instant::now();
        let mut progress = progress.add_child("Traverse commit graph");
        progress.init(None, progress::count("commits"));
        let mut commits: Vec<Vec<u8>> = Vec::new();
        let commit_iter = interrupt::Iter::new(
            commit_id.ancestors(|oid, buf| {
                progress.inc();
                repo.objects.find(oid, buf).map(|o| {
                    commits.push(o.data.to_owned());
                    objs::CommitRefIter::from_bytes(o.data)
                })
            }),
            || anyhow!("Cancelled by user"),
        );
        let mut is_shallow = false;
        for c in commit_iter {
            match c? {
                Ok(c) => c,
                Err(git::traverse::commit::ancestors::Error::FindExisting { .. }) => {
                    is_shallow = true;
                    break;
                }
                Err(err) => return Err(err.into()),
            };
        }
        progress.show_throughput(start);
        (commits, is_shallow)
    };

    let mailmap = repo.load_mailmap();
    let start = Instant::now();
    #[allow(clippy::redundant_closure)]
    let mut all_commits: Vec<actor::Signature> = all_commits
        .into_par_iter()
        .filter_map(|commit_data: Vec<u8>| {
            objs::CommitRefIter::from_bytes(&commit_data)
                .author()
                .map(|author| mailmap.resolve(author.trim()))
                .ok()
        })
        .collect::<Vec<_>>();
    all_commits.sort_by(|a, b| {
        a.email.cmp(&b.email).then(
            a.time
                .seconds_since_unix_epoch
                .cmp(&b.time.seconds_since_unix_epoch)
                .reverse(),
        )
    });
    if all_commits.is_empty() {
        bail!("No commits to process");
    }
    let mut current_email = &all_commits[0].email;
    let mut slice_start = 0;
    let mut results_by_hours = Vec::new();
    for (idx, elm) in all_commits.iter().enumerate() {
        if elm.email != *current_email {
            results_by_hours.push(estimate_hours(&all_commits[slice_start..idx]));
            slice_start = idx;
            current_email = &elm.email;
        }
    }
    if let Some(commits) = all_commits.get(slice_start..) {
        results_by_hours.push(estimate_hours(commits));
    }

    let num_authors = results_by_hours.len();
    let mut results_by_hours = if !omit_unify_identities {
        deduplicate_identities(&results_by_hours)
    } else {
        results_by_hours
            .iter()
            .fold(Vec::with_capacity(results_by_hours.len()), |mut acc, e| {
                acc.push(e.into());
                acc
            })
    };
    let elapsed = start.elapsed();
    progress.done(format!(
        "Extracted and organized data from {} commits in {:?} ({:0.0} commits/s)",
        all_commits.len(),
        elapsed,
        all_commits.len() as f32 / elapsed.as_secs_f32()
    ));

    let num_unique_authors = results_by_hours.len();
    if show_pii {
        results_by_hours.sort_by(|a, b| a.hours.partial_cmp(&b.hours).unwrap_or(std::cmp::Ordering::Equal));
        for entry in results_by_hours.iter() {
            writeln!(out, "{}\n", entry)?;
        }
    }
    let (total_hours, total_commits) = results_by_hours
        .iter()
        .map(|e| (e.hours, e.num_commits))
        .reduce(|a, b| (a.0 + b.0, a.1 + b.1))
        .expect("at least one commit at this point");
    writeln!(
        out,
        "total hours: {:.02}\ntotal 8h days: {:.02}\ntotal commits = {}{}\ntotal authors: {}",
        total_hours,
        total_hours / HOURS_PER_WORKDAY,
        total_commits,
        is_shallow.then(|| " (shallow)").unwrap_or_default(),
        num_authors
    )?;
    if !omit_unify_identities {
        writeln!(
            out,
            "total unique authors: {} ({:.02}% duplication)",
            num_unique_authors,
            (1.0 - (num_unique_authors as f32 / num_authors as f32)) * 100.0
        )?;
    }
    assert_eq!(total_commits, all_commits.len() as u32, "need to get all commits");
    Ok(())
}

const MINUTES_PER_HOUR: f32 = 60.0;
const HOURS_PER_WORKDAY: f32 = 8.0;

fn estimate_hours(commits: &[actor::Signature]) -> WorkByEmail {
    assert!(!commits.is_empty());
    const MAX_COMMIT_DIFFERENCE_IN_MINUTES: f32 = 2.0 * MINUTES_PER_HOUR;
    const FIRST_COMMIT_ADDITION_IN_MINUTES: f32 = 2.0 * MINUTES_PER_HOUR;

    let hours = FIRST_COMMIT_ADDITION_IN_MINUTES / 60.0
        + commits.iter().rev().tuple_windows().fold(
            0_f32,
            |hours, (cur, next): (&actor::Signature, &actor::Signature)| {
                let change_in_minutes =
                    (next.time.seconds_since_unix_epoch - cur.time.seconds_since_unix_epoch) as f32 / MINUTES_PER_HOUR;
                if change_in_minutes < MAX_COMMIT_DIFFERENCE_IN_MINUTES {
                    hours + change_in_minutes as f32 / MINUTES_PER_HOUR
                } else {
                    hours + (FIRST_COMMIT_ADDITION_IN_MINUTES / MINUTES_PER_HOUR)
                }
            },
        );
    let author = &commits[0];
    WorkByEmail {
        name: author.name.to_owned(),
        email: author.email.to_owned(),
        hours,
        num_commits: commits.len() as u32,
    }
}

fn deduplicate_identities(persons: &[WorkByEmail]) -> Vec<WorkByPerson<'_>> {
    let mut email_to_index = HashMap::<&BString, usize>::with_capacity(persons.len());
    let mut name_to_index = HashMap::<&BString, usize>::with_capacity(persons.len());
    let mut out = Vec::<WorkByPerson<'_>>::with_capacity(persons.len());
    for person_by_email in persons {
        match email_to_index.entry(&person_by_email.email) {
            Entry::Occupied(email_entry) => {
                out[*email_entry.get()].merge(person_by_email);
                name_to_index.insert(&person_by_email.name, *email_entry.get());
            }
            Entry::Vacant(email_entry) => match name_to_index.entry(&person_by_email.name) {
                Entry::Occupied(name_entry) => {
                    out[*name_entry.get()].merge(person_by_email);
                    email_entry.insert(*name_entry.get());
                }
                Entry::Vacant(name_entry) => {
                    let idx = out.len();
                    name_entry.insert(idx);
                    email_entry.insert(idx);
                    out.push(person_by_email.into());
                }
            },
        }
    }
    out
}

#[derive(Debug)]
struct WorkByPerson<'a> {
    name: Vec<&'a BString>,
    email: Vec<&'a BString>,
    hours: f32,
    num_commits: u32,
}

impl<'a> WorkByPerson<'a> {
    fn merge(&mut self, other: &'a WorkByEmail) {
        if !self.name.contains(&&other.name) {
            self.name.push(&other.name);
        }
        if !self.email.contains(&&other.email) {
            self.email.push(&other.email);
        }
        self.num_commits += other.num_commits;
        self.hours += other.hours;
    }
}

impl<'a> From<&'a WorkByEmail> for WorkByPerson<'a> {
    fn from(w: &'a WorkByEmail) -> Self {
        WorkByPerson {
            name: vec![&w.name],
            email: vec![&w.email],
            hours: w.hours,
            num_commits: w.num_commits,
        }
    }
}

impl<'a> Display for WorkByPerson<'a> {
    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
        writeln!(f, "{} <{}>", self.name.iter().join(", "), self.email.iter().join(", "))?;
        writeln!(f, "{} commits found", self.num_commits)?;
        writeln!(
            f,
            "total time spent: {:.02}h ({:.02} 8h days)",
            self.hours,
            self.hours / HOURS_PER_WORKDAY
        )
    }
}

#[derive(Debug)]
struct WorkByEmail {
    name: BString,
    email: BString,
    hours: f32,
    num_commits: u32,
}