FMHYedit/.github/add-headers.py

45 lines
3.1 KiB
Python
Raw Normal View History

import os
headersData = {
"AdblockVPNGuide.md": [":name_badge:", "140", "# Adblocking / Privacy", "Adblocking, Privacy, VPN's, Proxies, Antivirus"],
"AI.md": [":robot_face:", "130", "# Artificial Intelligence", "Chat Bots, Text Generators, Image Generators, ChatGPT Tools"],
"Android-iOSGuide.md": [":iphone:", "45", "# Android / iOS", "Apps, Jailbreaking, Android Emulators"],
"AudioPiracyGuide.md": [":musical_note:", "110", "# Music / Podcasts / Radio", "Stream Audio, Download Audio, Torrent Audio"],
"Beginners-Guide.md": [":pirate_flag:", "150", "# Beginners Guide to Piracy", ""],
"DEVTools.md": [":male-technologist:", "50", "# Developer Tools", ""],
"DownloadPiracyGuide.md": [":floppy_disk:", "80", "# Downloading", "Download Sites, Software Sites, Open Directories"],
"EDUPiracyGuide.md": [":brain:", "60", "# Educational", "Courses, Documentaries, Learning Resources"],
"GamingPiracyGuide.md": [":video_game:", "100", "# Gaming / Emulation", "Download Games, ROMs, Gaming Tools"],
"LinuxGuide.md": [":penguin:", "40", "# Linux / MacOS", "Apps, Software Sites, Gaming"],
"MISCGuide.md": [":open_file_folder:", "30", "# Miscellaneous", "Extensions, Indexes, News, Health, Food, Fun"],
"NSFWPiracy.md": [":underage:", "25", "# NSFW", ""],
"Non-English.md": [":earth_asia:", "35", "# Non-English", "International Piracy Sites"],
"ReadingPiracyGuide.md": [":green_book:", "90", "# Books / Comics / Manga", "Books, Comics, Magazines, Newspapers"],
"STORAGE.md": [":card_file_box:", "1", "", ""],
"TOOLSGuide.md": [":wrench:", "58", "# Tools", "General Tools, Internet Tools, System Tools"],
"TorrentPiracyGuide.md": [":cyclone:", "70", "# Torrenting", "Torrent Clients, Torrent Sites, Trackers"],
"VideoPiracyGuide.md": [":tv:", "120", "# Movies / TV / Anime", "Stream Videos, Download Videos, Torrent Videos"],
"base64.md": [":key:", "20", "", ""],
"img-tools.md": [":camera:", "55", "# Image Tools", ""],
2023-06-19 06:30:00 +00:00
"UnsafeSites.md": [":warning:", "21", "# Unsafe Sites", ""]
}
def getHeaderForPage(pageFilename):
data = headersData[pageFilename]
header = '---\n' + 'icon: ' + '"' + data[0] + '"' + '\n' + 'order: ' + data[1] + '\n' + '---\n' + data[2] + '\n' + data[3] + '\n\n'
return header
def apply_to_all_md_files_in_current_dir():
files = os.listdir('.')
for file in files:
if file in headersData:
with open(file, 'r', encoding='utf-8') as f:
content = f.read()
if not content.startswith('---'):
print("adding header to " + file)
with open(file, 'w', encoding='utf-8') as f2:
header = getHeaderForPage(file)
f2.write(header+content)
2023-06-19 06:30:00 +00:00
apply_to_all_md_files_in_current_dir()