scrape_gel.py

import os
import re
import sys
import time
import utils
import random
import urllib
import asyncio
import aiohttp
import aiofiles
import argparse
import concurrent
import html as libhtml
from constants import *
from bs4 import BeautifulSoup
from collections import defaultdict

def get_tags(soup, preserve_meta_tags):
    tag_ul = soup.find("ul", id="tag-list")
    if not tag_ul:
        raise RuntimeError("No tag list found in this web page!")
    type_tag_list_dict = defaultdict(list)
    for element in tag_ul.find_all("li"):
        class_name = element.get("class")
        if not class_name or len(class_name) != 1:
            continue
        class_name = class_name[0]
        if not class_name.startswith("tag-type-"):
            continue
        type_tag_list_dict[class_name[9:]].append(element.find("a", recursive=False).contents[0].replace("_", " "))
    tags = []
    for type, tag_list in type_tag_list_dict.items():
        if not preserve_meta_tags:
            match type:
                case "copyright":
                    if "minecraft" in tag_list:
                        tags.append("minecraft")
                    continue
                case "artist":
                    continue
                case "metadata":
                    continue
        tags += tag_list
    return tags

async def process_link(scrape_args, scrape_state):
    image_id = re.search(r"id=(\d+)", scrape_args.target).group(1)
    image_id_int = int(image_id)
    scrape_state.last_reached_image_id = image_id_int
    image_id_already_exists = image_id in scrape_state.existing_image_ids
    if image_id_already_exists and image_id_int % 100 < 99:
        # print(f"Image {image_id} already exists, skipped.")
        return
    scrape_state.existing_image_ids.add(image_id)
    error = None
    for i in range(1, MAX_RETRY + 2): # 1 indexed.
        try:
            if utils.get_sigint_count() >= 1 or isinstance(scrape_args.max_scrape_count, int) and scrape_state.scraped_image_count >= scrape_args.max_scrape_count:
                break
            # print(f"Processing image {image_id}...")
            query_start_time = time.time()
            async with scrape_state.session.get(scrape_args.target) as response:
                html = await response.text()
            query_used_time = time.time() - query_start_time
            soup = BeautifulSoup(html, "html.parser")
            score_span = soup.find("span", id="psc" + image_id)
            if score_span:
                scrape_state.last_reached_image_score = int(score_span.contents[0])
            if image_id_already_exists:
                # print(f"Image {image_id} already exists, skipped.")
                return
            video_container = soup.find("video", id="gelcomVideoPlayer")
            if video_container:
                print(f"Image {image_id} is a video, skipped.")
                return
            image_container = soup.find("section", class_=["image-container", "note-container"])
            if not image_container:
                raise RuntimeError("No image container found.")
            if not scrape_args.use_low_quality:
                image_download_url = soup.find("a", string="Original image")["href"]
            else:
                image_download_url = image_container.find("img", id="image")["src"]
            image_ext = os.path.splitext(image_download_url)[1].lower()
            if image_ext not in IMAGE_EXT:
                print(f"Image {image_id} is not an image, skipped.")
                return
            tags = get_tags(soup, scrape_args.preserve_meta_tags)
            tag_count = len(tags)
            if tag_count < scrape_args.min_tags:
                # print(f"Image {image_id} doesn't have enough tags({tag_count} < {scrape_args.min_tags}), skipped.")
                return
            rating = image_container["data-rating"]
            if rating == "explicit": tags.append("nsfw")
            elif rating == "questionable": tags.append("qfw")
            else: tags.append("sfw")
            random.shuffle(tags)
            image_path = os.path.join(IMAGE_DIR, image_id + image_ext)
            tags_path = os.path.join(IMAGE_DIR, image_id + ".txt")
            tags_text = ", ".join(tags)
            download_start_time = time.time()
            async with scrape_state.session.get(image_download_url) as img_response:
                img_data = await img_response.read()
            download_used_time = time.time() - download_start_time
            async with aiofiles.open(image_path, "wb") as f:
                await f.write(img_data)
            async with aiofiles.open(tags_path, "w", encoding="utf8") as f:
                await f.write(tags_text)
            if not await utils.submit_validation(scrape_state.thread_pool, image_path, tags_path, scrape_args.width, scrape_args.height, scrape_args.convert_to_avif):
                scrape_state.existing_image_ids.remove(image_id)
            else:
                scrape_state.scraped_image_count += 1
                total_query_time = scrape_state.avg_query_time[0] * scrape_state.avg_query_time[1] + query_used_time
                total_download_time = scrape_state.avg_download_time[0] * scrape_state.avg_download_time[1] + download_used_time
                scrape_state.avg_query_time[1] += 1
                scrape_state.avg_download_time[1] += 1
                scrape_state.avg_query_time[0] = total_query_time / scrape_state.avg_query_time[1]
                scrape_state.avg_download_time[0] = total_download_time / scrape_state.avg_download_time[1]
                interval = 1000
                if scrape_state.scraped_image_count % interval == 0:
                    print(
                        f"Scraped {scrape_state.scraped_image_count}/{scrape_args.max_scrape_count} images, "
                        f"stats for the last {interval} images: [Average query time: {scrape_state.avg_query_time[0]:.3f}s | Average download time: {scrape_state.avg_download_time[0]:.3f}s]"
                    )
                    scrape_state.avg_query_time = [0.0, 0]
                    scrape_state.avg_download_time = [0.0, 0]
            return
        except Exception as e:
            error = e
            if i > MAX_RETRY:
                break
            # print(f"A {e.__class__.__name__} occurred with image {image_id}: {e}\nPausing for 0.1 second before retrying attempt {i}/{MAX_RETRY}...")
            await asyncio.sleep(0.1)
    if not image_id_already_exists:
        scrape_state.existing_image_ids.remove(image_id)
    print(f"All retry attempts failed, image {image_id} skipped. Final error {error.__class__.__name__}: {error}")

def parse_args():
    parser = argparse.ArgumentParser(description="Scrape images from Gelbooru.")
    parser.add_argument("-s", "--site", default="https://gelbooru.com", help="Domain to scrape from, default to https://gelbooru.com")
    parser.add_argument("-W", "--width", type=int, help="Scale the width of the image to the specified value, must either provide both width and height or not provide both")
    parser.add_argument("-H", "--height", type=int, help="Scale the height of the image to the specified value, must either provide both width and height or not provide both")
    parser.add_argument("-a", "--avif", action="store_true", help="If set, will convert the image into avif, need to have pillow-avif-plugin installed")
    parser.add_argument("-l", "--low-quality", action="store_true", help="If set, will download the sample instead of the original image")
    parser.add_argument("-t", "--min-tags", type=int, default=0, help="Filter out images with less than the specified amount of tags, default to 0")
    parser.add_argument("-p", "--preserve-meta-tags", action="store_true", help="Preserve artist, copyright, and metadata tags")
    parser.add_argument("-m", "--max-scrape-count", type=int, help="Stop after scraping the set amount of images, may not be exact because of the asynchronous nature of this script, default to infinite")
    parser.add_argument("-c", "--continuous-scraping", action="store_true", help="If set, will scraping continuously even when reaching the 20000 images Gelbooru search depth cap by adjusting search tags")
    parser.add_argument("tags_to_search", nargs=argparse.REMAINDER, help="List of tags to search for, default to all")
    args = parser.parse_args()
    if args.width is None or args.height is None:
        if args.width is not None or args.height is not None:
            print("You must either provide both width and height or not provide both at the same time!")
            sys.exit(1)
    else:
        if args.width < 1:
            print("Width must be greater than or equal to 1!")
            sys.exit(1)
        if args.height < 1:
            print("Height must be greater than or equal to 1!")
            sys.exit(1)
    if args.avif:
        try:
            import pillow_avif
        except ImportError:
            print("You need to pip install pillow-avif-plugin to use avif conversion!")
            sys.exit(1)
    if args.min_tags < 0:
        print("Minimum tags must be greater than or equal to 0!")
        sys.exit(1)
    if isinstance(args.max_scrape_count, int) and args.max_scrape_count <= 0:
        print("Maximum scrape count must be greater than 0!")
        sys.exit(1)
    if not args.tags_to_search:
        args.tags_to_search = ["all"]
    return args

async def main():
    args = parse_args()
    print("Starting...")
    page_number = 0
    search_tags = utils.SearchTags(args.tags_to_search)

    os.makedirs(IMAGE_DIR, exist_ok=True)
    existing_image_ids = utils.get_existing_image_id_set(IMAGE_DIR)
    utils.register_sigint_callback()

    session_args = [TIMEOUT, {"fringeBenefits": "yup"}]
    scrape_state = utils.ScrapeState(concurrent.futures.ThreadPoolExecutor(max_workers=os.cpu_count()), utils.get_session(*session_args), existing_image_ids)
    session_refresh_counter = 0
    tasks = []
    while True:
        try:
            if utils.get_sigint_count() >= 1 or isinstance(args.max_scrape_count, int) and scrape_state.scraped_image_count >= args.max_scrape_count:
                break
            request_url = f"{args.site}/index.php?page=post&s=list&tags={search_tags.to_search_string()}&pid={page_number}"
            print(f"Going to {request_url}")
            async with scrape_state.session.get(request_url) as response:
                html = await response.text()
            soup = BeautifulSoup(html, "html.parser")
            thumbnails_div = soup.find("div", class_="thumbnail-container")
            if not thumbnails_div:
                raise RuntimeError("Thumbnails division not found.")
            notice_error = thumbnails_div.find("div", class_="notice error")
            if notice_error and args.continuous_scraping:
                print("Reached restricted depth, adjusting search tags to continue scraping...")
                search_tags.update_bound(scrape_state)
                page_number = 0
                continue
            image_urls = [a["href"] for a in thumbnails_div.find_all("a")]
            image_url_count = len(image_urls)
            if image_url_count == 0:
                print("Website returned 0 image urls.")
                break
            print(f"Got {image_url_count} posts.")
            page_number += image_url_count
            for image_url in image_urls:
                if utils.get_sigint_count() >= 1 or isinstance(args.max_scrape_count, int) and scrape_state.scraped_image_count >= args.max_scrape_count:
                    break
                while len(tasks) >= MAX_TASKS:
                    if utils.get_sigint_count() >= 1 or isinstance(args.max_scrape_count, int) and scrape_state.scraped_image_count >= args.max_scrape_count:
                        break
                    await asyncio.sleep(0.1)
                    for i in range(len(tasks) - 1, -1, -1):
                        task = tasks[i]
                        if task.done():
                            await task
                            del tasks[i]
                tasks.append(asyncio.create_task(process_link(
                    utils.ScrapeArgs(image_url, args.width, args.height, args.avif, args.low_quality, args.min_tags, args.preserve_meta_tags, args.max_scrape_count), scrape_state
                )))
            if utils.get_sigint_count() >= 1 or isinstance(args.max_scrape_count, int) and scrape_state.scraped_image_count >= args.max_scrape_count:
                break
            session_refresh_counter += 1
            if session_refresh_counter % 50 == 0:
                print("Refreshing session...")
                while tasks and utils.get_sigint_count() < 1:
                    await asyncio.sleep(0.1)
                    for i in range(len(tasks) - 1, -1, -1):
                        task = tasks[i]
                        if task.done():
                            await task
                            del tasks[i]
                if utils.get_sigint_count() < 1:
                    await scrape_state.session.close()
                    scrape_state.session = utils.get_session(*session_args)
        except Exception as e:
            print(f"An error occurred: {e}\nPausing for 0.1 second before retrying...")
            await asyncio.sleep(0.1)
    if utils.get_sigint_count() >= 1:
        print("Script interrupted by user, gracefully exiting...\nYou can interrupt again to exit semi-forcefully, but it will break image checks!")
    else:
        print("No more images to download, waiting already submitted tasks to finish...")
    while tasks and utils.get_sigint_count() <= 1:
        await asyncio.sleep(0.1)
        for i in range(len(tasks) - 1, -1, -1):
            task = tasks[i]
            if task.done():
                await task
                del tasks[i]
    await scrape_state.session.close()
    if utils.get_sigint_count() >= 1:
        if utils.get_sigint_count() >= 2:
            print("Another interrupt received, exiting semi-forcefully...\nYou can interrupt again for truly forceful exit, but it most likely will break a lot of things!")
        sys.exit(1)

if __name__ == "__main__":
    try:
        asyncio.run(main())
    except KeyboardInterrupt:
        print("\nScript interrupted by user, exiting...")
        sys.exit(1)