noaa-goes-visualization/puller.py

import os
import urllib.request
import urllib.parse
import re
import time
import random
import datetime
import json
from threading import Thread
import queue

import tqdm

def recursive_find_links(url):
    links_regex_pattern = r'(?<=<a href=")([^ ?]*)(?=">)' # Find href links that do not contain question marks or whitespace
    times_regex_pattern = r'(?<=<\/a>)(\s*\d{4}-\d{2}-\d{2} \d{2}:\d{2})(?= )' # Find timestamps in UTC in the format YYYY-MM-DD HH:mm
    with urllib.request.urlopen(url) as response:
        html_content = response.read().decode('utf-8')

        links = re.findall(links_regex_pattern, html_content)
        times = re.findall(times_regex_pattern, html_content)
        for i in range(len(times)):
            dt = datetime.datetime.strptime(times[i].strip(), "%Y-%m-%d %H:%M")
            dt.replace(tzinfo=datetime.timezone.utc)
            times[i] = time.mktime(dt.timetuple())
        
        if len(links) - 1 == len(times):
            links = links[1:]
        elif len(links) == len(times):
            pass
        else:
            raise(ValueError)

        for _link, _time in zip(links, times):
            if _link.endswith("latest.png"):
                continue
            elif _link.endswith("/"):
                yield from recursive_find_links(urllib.parse.urljoin(url,_link))
            else:
                yield urllib.parse.urljoin(url,_link), _time
    

# Fetch image from url and store it to path, retrying on failure
def image_fetch_worker(work_queue, result_queue, attempt_count = 1):
    while True:
        job = work_queue.get()
        if job == None:
            return
        url, path, t = job

        attempts = 0
        while attempts < attempt_count:
            try:
                req = urllib.request.Request(url, data=None)
                image_data = urllib.request.urlopen(req).read()
                os.makedirs(os.path.split(path)[0], exist_ok=True)
                open(path, 'wb').write(image_data)
                result_queue.put((True, url, t))
                break
            except Exception as e:
                if hasattr(e, "code") and e.code == 404: # This is expected if the file has been removed from the site (at least for swpc.noaa.gov)
                    attempts += attempt_count
                elif attempts == 0:
                    print(f"\nA problem occurred on image: {url} | {e}")
                    time.sleep(1 + random.random())
                    attempts += 1
            if (attempt_count > 1) and (attempts == attempt_count):
                print(f'\nAfter {attempt_count} retries, could not fetch: {url}')
                result_queue.put((False, url, t))
                break


if __name__ == "__main__":
    directory_url = r"https://services.swpc.noaa.gov/images/animations/suvi/"
    stored_images_dir = r"..\suvi"
    file_database_path = r"..\file_database.json"
    fetch_interval = 30*60
    nworkers = 8

    file_info_cache = {}
    try:
        print(f"Attempting to load file records from cache: {file_database_path}")
        with open(file_database_path, 'r') as f:
            file_info_cache = json.loads(f.read())
        print(f"File records loaded from cache: {len(file_info_cache)} records found.")
    except Exception as e:
        print(f"Load failed, starting with empty cache")
        file_info_cache = {}

    work_queue = queue.Queue(maxsize=nworkers)
    result_queue = queue.Queue()
    workers = []
    for _ in range(nworkers):
        t = Thread(target=image_fetch_worker, args=(work_queue, result_queue), daemon=True)
        t.start()
        workers.append(t)
    try:
        while True:
            fetched_image_count = 0
            already_had_image_count = 0
            failed_image_count = 0
            urllen = len(directory_url)
            for l, t in tqdm.tqdm(recursive_find_links(directory_url), desc="Downloading files"):
                # Collect complete work and record it
                while True:
                    try:
                        r_success, r_url, r_t = result_queue.get_nowait()
                        if r_success: 
                            fetched_image_count += 1
                            file_info_cache[r_url] = r_t
                        else:
                            failed_image_count += 1
                    except queue.Empty:
                        break
                # If we dont have the file or the file at the link is newer than the one we previously fetched
                if (not (l in file_info_cache)) or t > file_info_cache[l]: 
                    file_portion_of_link = l[urllen:]
                    filepath = os.path.join(stored_images_dir, file_portion_of_link.replace("/", os.sep))
                    if os.path.exists(filepath):
                        if l in file_info_cache: # If we have record of this file, it must be out of date, rename it and download the new version.
                            os.rename(filepath, filepath + f"_{int(file_info_cache[l])}")
                        else: # If we have no record of this file, update the file info cache and don't redownload
                            file_info_cache[l] = t
                            already_had_image_count += 1
                            continue
                    work_queue.put((l, filepath, t))
                else:
                    already_had_image_count += 1
            with open(file_database_path, 'w') as f:
                f.write(json.dumps(file_info_cache))
            print(f"Downloaded {fetched_image_count} | Already had {already_had_image_count} | Failed {failed_image_count}")
            time.sleep(fetch_interval)
    except KeyboardInterrupt:
        print("Saving file database and shutting down.")

    for _ in range(nworkers):
        try:
            work_queue.put(None, timeout=1.0)
        except:
            break

    for w in workers:
        w.join(5.0)

    with open(file_database_path, 'w') as f:
        f.write(json.dumps(file_info_cache))

    print(f"Downloaded {fetched_image_count} | Already had {already_had_image_count} | Failed {failed_image_count}")