#!/usr/bin/env python3

#
# Sends updates to a Discord webhook for new changelog entries since the last GitHub Actions publish run.
# Automatically figures out the last run and changelog contents with the GitHub API.
#

import io
import itertools
import os
import requests
import yaml
from typing import Any, Iterable

GITHUB_API_URL    = os.environ.get("GITHUB_API_URL", "https://api.github.com")
GITHUB_REPOSITORY = os.environ["GITHUB_REPOSITORY"]
GITHUB_RUN        = os.environ["GITHUB_RUN_ID"]
GITHUB_TOKEN      = os.environ["GITHUB_TOKEN"]

# https://discord.com/developers/docs/resources/webhook
DISCORD_SPLIT_LIMIT = 2000
DISCORD_WEBHOOK_URL = os.environ.get("DISCORD_WEBHOOK_URL")

CHANGELOG_FILE = "Resources/Changelog/Changelog.yml"

TYPES_TO_EMOJI = {
    "Fix":    "🐛",
    "Add":    "🆕",
    "Remove": "❌",
    "Tweak":  "⚒️"
}

ChangelogEntry = dict[str, Any]

def main():
    if not DISCORD_WEBHOOK_URL:
        return

    session = requests.Session()
    session.headers["Authorization"]        = f"Bearer {GITHUB_TOKEN}"
    session.headers["Accept"]               = "Accept: application/vnd.github+json"
    session.headers["X-GitHub-Api-Version"] = "2022-11-28"

    most_recent = get_most_recent_workflow(session)
    last_sha = most_recent['head_commit']['id']
    print(f"Last successful publish job was {most_recent['id']}: {last_sha}")
    last_changelog = yaml.safe_load(get_last_changelog(session, last_sha))
    with open(CHANGELOG_FILE, "r") as f:
        cur_changelog = yaml.safe_load(f)

    diff = diff_changelog(last_changelog, cur_changelog)
    send_to_discord(diff)


def get_most_recent_workflow(sess: requests.Session) -> Any:
    workflow_run = get_current_run(sess)
    past_runs = get_past_runs(sess, workflow_run)
    for run in past_runs['workflow_runs']:
        # First past successful run that isn't our current run.
        if run["id"] == workflow_run["id"]:
            continue

        return run


def get_current_run(sess: requests.Session) -> Any:
    resp = sess.get(f"{GITHUB_API_URL}/repos/{GITHUB_REPOSITORY}/actions/runs/{GITHUB_RUN}")
    resp.raise_for_status()
    return resp.json()


def get_past_runs(sess: requests.Session, current_run: Any) -> Any:
    """
    Get all successful workflow runs before our current one.
    """
    params = {
        "status": "success",
        "created": f"<={current_run['created_at']}"
    }
    resp = sess.get(f"{current_run['workflow_url']}/runs", params=params)
    resp.raise_for_status()
    return resp.json()


def get_last_changelog(sess: requests.Session, sha: str) -> str:
    """
    Use GitHub API to get the previous version of the changelog YAML (Actions builds are fetched with a shallow clone)
    """
    params = {
        "ref": sha,
    }
    headers = {
        "Accept": "application/vnd.github.raw"
    }

    resp = sess.get(f"{GITHUB_API_URL}/repos/{GITHUB_REPOSITORY}/contents/{CHANGELOG_FILE}", headers=headers, params=params)
    resp.raise_for_status()
    return resp.text


def diff_changelog(old: dict[str, Any], cur: dict[str, Any]) -> Iterable[ChangelogEntry]:
    """
    Find all new entries not present in the previous publish.
    """
    old_entry_ids = {e["id"] for e in old["Entries"]}
    return (e for e in cur["Entries"] if e["id"] not in old_entry_ids)


def get_discord_body(content: str):
    return {
            "content": content,
            # Do not allow any mentions.
            "allowed_mentions": {
                "parse": []
            },
            # SUPPRESS_EMBEDS
            "flags": 1 << 2
        }


def send_discord(content: str):
    body = get_discord_body(content)

    response = requests.post(DISCORD_WEBHOOK_URL, json=body)
    response.raise_for_status()


def send_to_discord(entries: Iterable[ChangelogEntry]) -> None:
    if not DISCORD_WEBHOOK_URL:
        print(f"No discord webhook URL found, skipping discord send")
        return

    message_content = io.StringIO()
    # We need to manually split messages to avoid discord's character limit
    # With that being said this isn't entirely robust
    # e.g. a sufficiently large CL breaks it, but that's a future problem

    for name, group in itertools.groupby(entries, lambda x: x["author"]):
        # Need to split text to avoid discord character limit
        group_content = io.StringIO()
        group_content.write(f"**{name}** updated:\n")

        for entry in group:
            for change in entry["changes"]:
                emoji = TYPES_TO_EMOJI.get(change['type'], "❓")
                message = change['message']
                url = entry.get("url")
                if url and url.strip():
                    group_content.write(f"{emoji} [-]({url}) {message}\n")
                else:
                    group_content.write(f"{emoji} - {message}\n")

        group_text = group_content.getvalue()
        message_text = message_content.getvalue()
        message_length = len(message_text)
        group_length = len(group_text)

        # If adding the text would bring it over the group limit then send the message and start a new one
        if message_length + group_length >= DISCORD_SPLIT_LIMIT:
            print("Split changelog  and sending to discord")
            send_discord(message_text)

            # Reset the message
            message_content = io.StringIO()

        # Flush the group to the message
        message_content.write(group_text)
    
    # Clean up anything remaining
    message_text = message_content.getvalue()
    if len(message_text) > 0:
        print("Sending final changelog to discord")
        send_discord(message_text)


main()