svr/usr/bin/dict-maker

#!/usr/bin/env python3
import os
import re
import subprocess
import sys
from collections import defaultdict
from pathlib import Path
import argparse
import re
from typing import Pattern

REGEX: Pattern[str] = re.compile(r"...")

# ---- ensure spacy and model ----
try:
    import spacy
except ImportError:
    subprocess.check_call([sys.executable, "-m", "pip", "install", "spacy"])
    import spacy

def load_model(name="fr_core_news_sm"):
    try:
        return spacy.load(name)
    except OSError:
        subprocess.check_call([sys.executable, "-m", "spacy", "download", name])
        return spacy.load(name)

# ---- args ----
argparser = argparse.ArgumentParser()
argparser.add_argument("-v","--vault", default=".", help="Path to Obsidian vault")
argparser.add_argument("-d","--dict", default="Dictionary", help="Name of the dictionary directory")
args = argparser.parse_args()

# ---- config ----
VAULT_DIR = Path(args.vault)
DICT_DIR = VAULT_DIR / args.dict
WORD_REGEX = re.compile(r"\b[a-zA-Z]{3,}\b")

nlp = load_model("fr_core_news_sm")

# ---- prep ----
DICT_DIR.mkdir(exist_ok=True)
lemma_map = defaultdict(lambda: {"forms": set(), "files": set()})

# ---- scan ----

nlp.Defaults.stop_words.add(os.path.basename(os.getcwd()))
nlp.Defaults.stop_words.add("author")
nlp.Defaults.stop_words.add("jpeg")
nlp.Defaults.stop_words.add("jpg")
nlp.Defaults.stop_words.add("post")
nlp.Defaults.stop_words.add("like")
nlp.Defaults.stop_words.add("likes")
nlp.Defaults.stop_words.add("repost")
nlp.Defaults.stop_words.add("avatar")
nlp.Defaults.stop_words.add("bsky")
nlp.Defaults.stop_words.add("media")
nlp.Defaults.stop_words.add("thumnail")
nlp.Defaults.stop_words.add("http")
nlp.Defaults.stop_words.add("https")
nlp.Defaults.stop_words.add("com")
nlp.Defaults.stop_words.add("followers")
nlp.Defaults.stop_words.add("following")
nlp.Defaults.stop_words.add("unknown")
nlp.Defaults.stop_words.add("date")
nlp.Defaults.stop_words.add("social")
nlp.Defaults.stop_words.add("thumbnail")
nlp.Defaults.stop_words.add("replier")
nlp.Defaults.stop_words.add("replie")
nlp.Defaults.stop_words.add("for")
nlp.Defaults.stop_words.add("you")
nlp.Defaults.stop_words.add("from")
nlp.Defaults.stop_words.add("to")
nlp.Defaults.stop_words.add("script")
nlp.Defaults.stop_words.add("grep")
nlp.Defaults.stop_words.add("localhost")
nlp.Defaults.stop_words.add("sudo")
nlp.Defaults.stop_words.add("not")

for md_file in VAULT_DIR.rglob("*.md"):
    if "Dictionary" in md_file.parts:
        continue
    text = md_file.read_text(encoding="utf-8", errors="ignore")
    words = WORD_REGEX.findall(text.lower())
    doc = nlp(" ".join(words))
    for token in doc:
        if token.is_stop:
            continue
        lemma = token.lemma_
        if lemma.isalpha() and lemma not in nlp.Defaults.stop_words:
            lemma_map[lemma]["forms"].add(token.text)
            lemma_map[lemma]["files"].add(md_file)

print(f"Found {len(lemma_map)} lemmas.")

# ---- write ----
for lemma, data in lemma_map.items():
    if len(data["files"]) < 2:
        continue
    file_path = DICT_DIR / f"{lemma}.md"
    with open(file_path, "w", encoding="utf-8") as f:
        f.write(f"# {lemma}\n\n---\n")
        f.write(f"lemma: {lemma}\n")
        f.write(f"forms: [{', '.join(sorted(data['forms']))}]\n---\n\n")
        f.write("## Forms\n")
        for form in sorted(data["forms"]):
            f.write(f"- {form}\n")
        f.write("\n## Found in\n")
        for md in sorted(data["files"]):
            rel_path = md.relative_to(VAULT_DIR)
            f.write(f"- [[{rel_path}]]\n")


print(f"Dictionary generated in {DICT_DIR}")