"""Canonical full-text TextGrid subset reader/writer. No Praat or forced alignment run.""" import json,re,math,copy,hashlib def number(x): if type(x)not in [int,float]or not math.isfinite(x):raise ValueError('finite authored number') s=str(x) if type(x)is int else repr(x) return s[:-2] if s.endswith('.0') else s def quoted(s): if not isinstance(s,str)or any(c in s for c in ['\n','\r']):raise ValueError('single-line string subset') return '"'+s.replace('"','""')+'"' def validate(g): lo,hi=g['xmin'],g['xmax'];number(lo);number(hi) if lo>=hi or len(g['tiers'])!=2:raise ValueError('positive grid and exactly two declared tiers') for i,t in enumerate(g['tiers']): if not isinstance(t['entries'],list)or len(t['entries'])>100:raise ValueError('bounded authored entry count, at most100 before writing or reading') if t['xmin']!=lo or t['xmax']!=hi:raise ValueError('strict local tier coverage, not universal Praat validity rule') if t['class']!=['IntervalTier','TextTier'][i]:raise ValueError('one interval followed by one point tier subset') quoted(t['name']) if i==0: if not t['entries']or t['entries'][0]['xmin']!=lo or t['entries'][-1]['xmax']!=hi:raise ValueError('strict local interval coverage') last=lo for x in t['entries']: number(x['xmin']);number(x['xmax']);quoted(x['text']) if x['xmin']!=last or x['xmin']>=x['xmax']:raise ValueError('positive adjacent intervals required') last=x['xmax'] else: last=None for x in t['entries']: number(x['number']);quoted(x['mark']) if not lo<=x['number']<=hi or last is not None and x['number']<=last:raise ValueError('strictly ordered points within local domain') last=x['number'] def serialize(g): validate(g);lines=['File type = "ooTextFile"','Object class = "TextGrid"','',f"xmin = {number(g['xmin'])}",f"xmax = {number(g['xmax'])}",'tiers? ','size = 2','item []:'] for i,t in enumerate(g['tiers'],1): lines += [f' item [{i}]:',f" class = {quoted(t['class'])}",f" name = {quoted(t['name'])}",f" xmin = {number(t['xmin'])}",f" xmax = {number(t['xmax'])}"] label='intervals'if i==1 else'points';lines.append(f" {label}: size = {len(t['entries'])}") for j,x in enumerate(t['entries'],1): lines.append(f' {label} [{j}]:') for k in (['xmin','xmax','text']if i==1 else['number','mark']):lines.append(' '+k+' = '+(quoted(x[k])if k in ['text','mark']else number(x[k]))) return '\n'.join(lines)+'\n' def parse(text): # Labels and ordering are deliberately stricter than Praat's general reader. lines=[x.strip()for x in text.splitlines()if x.strip()];pos=0 def expect(value): nonlocal pos if pos>=len(lines)or lines[pos]!=value:raise ValueError('unsupported canonical full-text subset near '+value) pos+=1 def field(label,kind): nonlocal pos if pos>=len(lines)or not lines[pos].startswith(label+' = '):raise ValueError('missing '+label) raw=lines[pos][len(label)+3:];pos+=1 if kind=='str': if not re.fullmatch(r'"(?:""|[^"])*"',raw):raise ValueError('quoted string with doubled quotes required') return raw[1:-1].replace('""','"') if not re.fullmatch(r'-?(?:[0-9]+(?:\.[0-9]*)?|\.[0-9]+)(?:[eE][+-]?[0-9]+)?',raw):raise ValueError('finite numeric token required') x=float(raw);number(x) if kind=='int'and(x!=int(x)or not 0<=x<=100):raise ValueError('bounded authored entry count') return int(x)if kind=='int'else x expect('File type = "ooTextFile"');expect('Object class = "TextGrid"');g={'xmin':field('xmin','num'),'xmax':field('xmax','num'),'tiers':[]};expect('tiers? ');expect('size = 2');expect('item []:') for i in [1,2]: expect(f'item [{i}]:');t={'class':field('class','str'),'name':field('name','str'),'xmin':field('xmin','num'),'xmax':field('xmax','num'),'entries':[]};label='intervals'if i==1 else'points';n=field(label+': size','int') for j in range(1,n+1): expect(f'{label} [{j}]:');entry={k:field(k,'str'if k in ['text','mark']else'num')for k in (['xmin','xmax','text']if i==1 else['number','mark'])};t['entries'].append(entry) g['tiers'].append(t) if pos!=len(lines):raise ValueError('unexpected trailing records') validate(g);return g def demo(): g={'xmin':0,'xmax':2,'tiers':[{'class':'IntervalTier','name':'واژه','xmin':0,'xmax':2,'entries':[{'xmin':0,'xmax':1,'text':'سلام "دوست"'},{'xmin':1,'xmax':2,'text':'پایان'}]},{'class':'TextTier','name':'رخداد','xmin':0,'xmax':2,'entries':[{'number':0.25,'mark':'شروع'},{'number':1.5,'mark':'تأکید'}]}]} text=serialize(g);parsed=parse(text);assert parsed==g and serialize(parsed)==text and '"سلام ""دوست"""'in text assert text.encode('utf-8').decode('utf-8')==text precise=copy.deepcopy(g);precise['tiers'][1]['entries']=[{'number':0.25000000000004,'mark':'اول'},{'number':0.25000000000005,'mark':'دوم'}] assert parse(serialize(precise))==precise hundred=copy.deepcopy(g);hundred['tiers'][1]['entries']=[{'number':i/100,'mark':str(i)}for i in range(100)] assert parse(serialize(hundred))==hundred invalid={} for name,mutate in [('zero_interval',lambda x:x['tiers'][0]['entries'][0].update(xmax=0)),('gap',lambda x:x['tiers'][0]['entries'][1].update(xmin=1.1)),('unordered_points',lambda x:x['tiers'][1]['entries'].reverse()),('strict_local_tier_coverage',lambda x:x['tiers'][1].update(xmax=1.9)),('over100_points',lambda x:x['tiers'][1].update(entries=[{'number':i/100,'mark':str(i)}for i in range(101)]))]: other=copy.deepcopy(g);mutate(other) try:serialize(other) except ValueError as e:invalid[name]=str(e) else:raise AssertionError('mutation accepted') short='File type = "ooTextFile"\nObject class = "TextGrid"\n0\n2\n\n0\n' try:parse(short) except ValueError:short_unsupported=True else:raise AssertionError('short format accepted by full-only reader') return {'authored_reference':g,'full_textgrid_utf8':text,'utf8_sha256':hashlib.sha256(text.encode()).hexdigest(),'parsed':parsed,'roundtrip_exact':True,'close_distinct_float_roundtrip':precise,'hundred_points_roundtrip':True,'invalid_local_contract_cases':invalid,'short_format_unsupported_not_inherently_corrupt':short_unsupported,'praat_or_audio_alignment_executed':False,'scope':'Canonical full-text subset only, exactly IntervalTier then TextTier, single-line labels, strict local grid/tier/interval coverage, at most100entries per tier on both write/read, repr roundtrip-safe floats. General Praat ignores many labels and permits other full/short/intermediate formats and weak-invariant violations; our rejection is not a universal corruption verdict. UTF16 reader not implemented.'} if __name__=='__main__':print(json.dumps(demo(),ensure_ascii=False,indent=2))