"""Apply src/normalize.py to all six tables (train s1,s2,s3 + test t1,t2,t3) with multiprocessing and attach transliterated names using the learned lexicons. Writes ROOT/norm/.parquet. Usage on the box: cd /data/amlc/code && ROOT=/data/amlc WORKERS=16 /data/venv/bin/python scripts/preprocess_normalize.py [tables...] """ import os, sys, json, time, re, unicodedata from multiprocessing import Pool import pyarrow as pa, pyarrow.parquet as pq HERE=os.path.dirname(os.path.dirname(os.path.abspath(__file__))); sys.path.insert(0, HERE) from src.normalize import normalize_record R=os.environ.get('ROOT','/data/amlc'); W=int(os.environ.get('WORKERS','16')); CH=int(os.environ.get('CHUNK','40000')) INDIC=re.compile(r'[ऀ-෿]') NAME_LEX={}; ADDR_LEX={} LEGAL_LATIN={'private','limited','pvt','ltd','llp','opc','inc','incorporated','corp','corporation','co','company','llc','plc','pc','lp','sa','sarl','sas','sasu','eurl','sci','ei','ets'} def _load_lex(): global NAME_LEX, ADDR_LEX p=f'{R}/artifacts/translit_name_lexicon.json'; q=f'{R}/artifacts/translit_addr_lexicon.json' if os.path.exists(p): NAME_LEX={k:v['latin'] for k,v in json.load(open(p)).items()} if os.path.exists(q): ADDR_LEX={k:v['latin'] for k,v in json.load(open(q)).items()} def _translit_tokens(raw_name): """Map Indic-script tokens of the RAW name through the lexicon (lexicon keys are raw whitespace tokens).""" out=[]; unk=0 for t in raw_name.split(): if INDIC.search(t): m=NAME_LEX.get(t) or NAME_LEX.get(t.strip('.,()')) if m: out.append(m) else: unk+=1; out.append(t) else: t2=unicodedata.normalize('NFKD',t); t2=''.join(c for c in t2 if not unicodedata.combining(c)).lower() t2=re.sub(r'[^a-z0-9]+',' ',t2).strip() if t2: out.extend(t2.split()) return out, unk def work(args): ids, names, addrs, countries = args if not NAME_LEX: _load_lex() rows=[]; errs=0 for i,(eid,n,a,c) in enumerate(zip(ids,names,addrs,countries)): try: r=normalize_record(n or '', a or '', c or '') except Exception as e: errs+=1; r={'country':(c or '').lower(),'name_core':'','name_legal':'','name_tokens':[],'name_spaceless':'','name_alt_names':[],'name_is_domain_form':False,'name_has_indic_script':bool(INDIC.search(n or '')),'name_full':'','addr_tokens':[],'addr_numbers':[],'addr_street_number':'','addr_city':'','addr_state':'','addr_parts':[],'addr_district':'','addr_localities':[],'addr_street':'','blocking_keys':[]} if r.get('name_has_indic_script'): toks,unk=_translit_tokens(n or ''); toks=[t for t in toks if t not in LEGAL_LATIN]; r['name_translit_tokens']=toks; r['name_translit_unknown']=unk else: r['name_translit_tokens']=r['name_tokens']; r['name_translit_unknown']=0 r['name_translit']=' '.join(r['name_translit_tokens']) # address: map Indic tokens (state/city names) through the address lexicon at=[] for t in r['addr_tokens']: if INDIC.search(t): at.append(ADDR_LEX.get(t, t)) else: at.append(t) r['addr_tokens_translit']=at r['entity_id']=eid; r['raw_name']=n or ''; r['raw_address']=a or '' r.pop('addr_postal',None) rows.append(r) return rows, errs SCHEMA=pa.schema([ ('entity_id',pa.string()),('country',pa.string()),('raw_name',pa.string()),('raw_address',pa.string()), ('name_core',pa.string()),('name_legal',pa.string()),('name_tokens',pa.list_(pa.string())),('name_spaceless',pa.string()), ('name_alt_names',pa.list_(pa.string())),('name_is_domain_form',pa.bool_()),('name_has_indic_script',pa.bool_()),('name_full',pa.string()), ('name_translit',pa.string()),('name_translit_tokens',pa.list_(pa.string())),('name_translit_unknown',pa.int32()), ('addr_tokens',pa.list_(pa.string())),('addr_tokens_translit',pa.list_(pa.string())),('addr_numbers',pa.list_(pa.string())), ('addr_street_number',pa.string()),('addr_city',pa.string()),('addr_state',pa.string()),('addr_parts',pa.list_(pa.string())), ('addr_district',pa.string()),('addr_localities',pa.list_(pa.string())),('addr_street',pa.string()),('blocking_keys',pa.list_(pa.string())), ]) def _to_table(rows): cols={f.name:[r.get(f.name) for r in rows] for f in SCHEMA} arrays=[] for f in SCHEMA: v=cols[f.name] if pa.types.is_list(f.type): v=[[str(x) for x in (vv or [])] for vv in v] elif pa.types.is_string(f.type): v=['' if vv is None else str(vv) for vv in v] elif pa.types.is_boolean(f.type): v=[bool(vv) for vv in v] else: v=[int(vv or 0) for vv in v] # Build the array with an explicit type so an all-empty list column (e.g. name_alt_names with # no rows matched in this chunk) is still typed list, not the list pyarrow infers # from an untyped Python list of empty lists — that mismatch broke ParquetWriter.write_table # across chunks with a fixed schema. arrays.append(pa.array(v, type=f.type)) return pa.Table.from_arrays(arrays, schema=SCHEMA) def run_table(tbl): t0=time.time(); src=f'{R}/parquet/{tbl}.parquet'; os.makedirs(f'{R}/norm',exist_ok=True); out=f'{R}/norm/{tbl}.parquet' T=pq.read_table(src, columns=['entity_id','business_name','business_address','country']) ids=T.column('entity_id').to_pylist(); names=T.column('business_name').to_pylist(); addrs=T.column('business_address').to_pylist(); cs=T.column('country').to_pylist() n=len(ids); chunks=[(ids[i:i+CH],names[i:i+CH],addrs[i:i+CH],cs[i:i+CH]) for i in range(0,n,CH)] print(f"[{tbl}] rows={n} chunks={len(chunks)} workers={W}", flush=True) writer=None; errs=0; done=0 with Pool(W) as pool: for rows,e in pool.imap(work, chunks, chunksize=1): errs+=e; done+=len(rows) tab=_to_table(rows) if writer is None: writer=pq.ParquetWriter(out, SCHEMA, compression='zstd') writer.write_table(tab) if done % (CH*25) == 0: print(f" {tbl}: {done}/{n} {time.time()-t0:.0f}s", flush=True) writer.close() print(f"[{tbl}] done rows={done} errors={errs} {time.time()-t0:.0f}s -> {out}", flush=True) if __name__=='__main__': tables=sys.argv[1:] or ['s1','s2','s3','t1','t2','t3'] _load_lex(); print("lexicons:", len(NAME_LEX), "name tokens,", len(ADDR_LEX), "address tokens", flush=True) for t in tables: run_table(t) print("ALL DONE", flush=True)