skrutable 2.6.1__tar.gz → 2.6.3__tar.gz

{skrutable-2.6.1 → skrutable-2.6.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skrutable
-Version: 2.6.1
+Version: 2.6.3
 Summary: skrutable library for working with Sanskrit text
 Home-page: https://github.com/tylergneill/skrutable
 Author: Tyler Neill

skrutable-2.6.3/src/skrutable/__init__.py

@@ -0,0 +1 @@
+__version__ = "2.6.3"

{skrutable-2.6.1 → skrutable-2.6.3}/src/skrutable/meter_identification.py

@@ -5,9 +5,13 @@ from skrutable.utils import _DEBUG_TIMING, _section_totals, timed
 import re
 import time as _time
 from copy import copy
+from concurrent.futures import ProcessPoolExecutor
 from dataclasses import dataclass
 from typing import Optional
 
+BATCH_MAX_WORKERS = 5
+BATCH_PARALLEL_THRESHOLD = 100
+
 # load config variables
 config = load_config_dict_from_json_file()
 scansion_syllable_separator = config["scansion_syllable_separator"]  # e.g. " "
@@ -50,10 +54,12 @@ def _verse_is_perfect(V):
 	return getattr(V, 'is_perfect', False)
 
 
-def flush_profiling_report(write_file=False):
+def flush_profiling_report(write_file=False, wall_clock_secs=None, parallel_workers=None):
 	"""Print the accumulated profiling table to stderr, then reset all counters.
 
 	Pass write_file=True to also write the table to profiling_debug.txt alongside the library source.
+	Pass wall_clock_secs to append a timing footer line.
+	Pass parallel_workers (int) to show worker count and parallelization speedup; omit or pass None for serial runs.
 	Safe to call even when _DEBUG_TIMING is False (no-op).
 	"""
 	if not _DEBUG_TIMING or not _category_totals:
@@ -124,6 +130,13 @@ def flush_profiling_report(write_file=False):
 		+ f'{total_scan:.2f}s'.rjust(sub_w)
 		+ f'{total_types:.2f}s'.rjust(sub_w)
 		+ '  ' + fmt_row(total_scan_vals, total_type_vals))
+	if wall_clock_secs is not None:
+		table_total = total_scan + total_types
+		if parallel_workers is not None:
+			speedup = table_total / wall_clock_secs if wall_clock_secs > 0 else float('inf')
+			lines.append(f'\n  table total (CPU across {parallel_workers} workers, inflated by overhead): {table_total:.2f}s  |  wall-clock: {wall_clock_secs:.2f}s  |  apparent parallelization speedup: {speedup:.2f}x')
+		else:
+			lines.append(f'\n  table total: {table_total:.2f}s  |  wall-clock: {wall_clock_secs:.2f}s')
 	block = '\n'.join(lines) + '\n'
 	if write_file:
 		timing_path = os.path.join(os.path.dirname(__file__), 'profiling_debug.txt')
@@ -140,6 +153,8 @@ class Diagnostic:
 	imperfect_label_sanskrit: Optional[dict] = None # keyed by pada (1–4 or 'odd'/'even'); Sanskrit only
 	imperfect_label_english: Optional[dict] = None  # keyed by pada (1–4 or 'odd'/'even'); English only
 	problem_syllables: Optional[dict] = None        # keyed by pada (1–4 or 'odd'/'even'); None if perfect
+	notable_syllables: Optional[dict] = None        # keyed by pada (1–4 or 'odd'/'even'); green-highlighted "interesting/ok" syllables
+	notable_label: Optional[dict] = None            # keyed by pada (1–4 or 'odd'/'even'); label for the notable feature (same string for skt/eng)
 
 	def perfect(self):
 		return self.perfect_id_label is not None
@@ -460,17 +475,28 @@ class VerseTester(object):
 				result = None
 				for weights_pattern, label in meter_patterns.anuzwuB_pAda['odd'].items():
 					if re.match(weights_pattern, odd_pAda_weights):
-						result = Diagnostic(perfect_id_label=label)
+						is_vipula = 'vipulā' in label
+						result = Diagnostic(
+							perfect_id_label=label,
+							notable_syllables={'odd': [4, 5, 6]} if is_vipula else None,
+							notable_label={'odd': label} if is_vipula else None,
+						)
 						break
 				if result is None:
 					# Odd pāda matched no perfect pattern — try asamīcīna patterns
 					# before falling back to the generic ya-gaṇa violation label.
 					for weights_pattern, (label, problem_syls, code) in meter_patterns.anuzwuB_pAda_asamIcIna['odd'].items():
 						if re.match(weights_pattern, odd_pAda_weights):
+							is_vipula = 'vipulā' in label
+							# extract vipulā name from label like "asamīcīnā, ma-vipulāyāḥ pūrvam raḥ syāt"
+							vipula_match = re.search(r'\w+-vipulā', label)
+							vipula_name = vipula_match.group(0) if vipula_match else None
 							result = Diagnostic(
 								imperfect_label_sanskrit={'odd': label},
 								imperfect_label_english={'odd': code},
 								problem_syllables={'odd': problem_syls},
+								notable_syllables={'odd': [4, 5, 6]} if is_vipula else None,
+								notable_label={'odd': vipula_name} if vipula_name else None,
 							)
 							break
 				if result is None:
@@ -515,7 +541,7 @@ class VerseTester(object):
 				Vrs.diagnostic = ardham_eva_result
 				return ardham_eva_result
 			elif ardham_eva_result.imperfect():
-				label_str = '; '.join(f"{k}: {v}" for k, v in ardham_eva_result.imperfect_label_sanskrit.items())
+				label_str = '; '.join(ardham_eva_result.imperfect_label_sanskrit.values())
 				Vrs.meter_label = f"anuṣṭubh (ardham eva: {label_str})"
 				Vrs.identification_score = meter_scores["anuṣṭubh, half, single half imperfect)"]
 				Vrs.is_perfect = False
@@ -540,14 +566,14 @@ class VerseTester(object):
 		# one half imperfect
 
 		elif pAdas_ab_result.imperfect() and pAdas_cd_result.perfect():
-			ab_str = '; '.join(f"{k}: {v}" for k, v in pAdas_ab_result.imperfect_label_sanskrit.items())
+			ab_str = '; '.join(pAdas_ab_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: {ab_str}; 3,4: {pAdas_cd_result.perfect_id_label})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, one half perfect, one imperfect)"]
 			Vrs.is_perfect = False
 			Vrs.diagnostic = {'ab': pAdas_ab_result, 'cd': pAdas_cd_result}
 			return pAdas_ab_result
 		elif pAdas_ab_result.perfect() and pAdas_cd_result.imperfect():
-			cd_str = '; '.join(f"{k}: {v}" for k, v in pAdas_cd_result.imperfect_label_sanskrit.items())
+			cd_str = '; '.join(pAdas_cd_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: {pAdas_ab_result.perfect_id_label}; 3,4: {cd_str})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, one half perfect, one imperfect)"]
 			Vrs.is_perfect = False
@@ -557,8 +583,8 @@ class VerseTester(object):
 		# both halves imperfect
 
 		elif pAdas_ab_result.imperfect() and pAdas_cd_result.imperfect():
-			ab_str = '; '.join(f"{k}: {v}" for k, v in pAdas_ab_result.imperfect_label_sanskrit.items())
-			cd_str = '; '.join(f"{k}: {v}" for k, v in pAdas_cd_result.imperfect_label_sanskrit.items())
+			ab_str = '; '.join(pAdas_ab_result.imperfect_label_sanskrit.values())
+			cd_str = '; '.join(pAdas_cd_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: {ab_str}; 3,4: {cd_str})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, both halves imperfect)"]
 			Vrs.is_perfect = False
@@ -568,14 +594,14 @@ class VerseTester(object):
 		# one half perfect, one length error
 
 		elif pAdas_ab_result.length_error() and pAdas_cd_result.perfect():
-			ab_str = '; '.join(f"{k}: {v}" for k, v in pAdas_ab_result.imperfect_label_sanskrit.items())
+			ab_str = '; '.join(pAdas_ab_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: ?? {ab_str}; 3,4: {pAdas_cd_result.perfect_id_label})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, one half perfect, one length error)"]
 			Vrs.is_perfect = False
 			Vrs.diagnostic = {'ab': pAdas_ab_result, 'cd': pAdas_cd_result}
 			return pAdas_cd_result
 		elif pAdas_ab_result.perfect() and pAdas_cd_result.length_error():
-			cd_str = '; '.join(f"{k}: {v}" for k, v in pAdas_cd_result.imperfect_label_sanskrit.items())
+			cd_str = '; '.join(pAdas_cd_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: {pAdas_ab_result.perfect_id_label}; 3,4: ?? {cd_str})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, one half perfect, one length error)"]
 			Vrs.is_perfect = False
@@ -585,16 +611,16 @@ class VerseTester(object):
 		# one half imperfect, one length error
 
 		elif pAdas_ab_result.length_error() and pAdas_cd_result.imperfect():
-			ab_str = '; '.join(f"{k}: {v}" for k, v in pAdas_ab_result.imperfect_label_sanskrit.items())
-			cd_str = '; '.join(f"{k}: {v}" for k, v in pAdas_cd_result.imperfect_label_sanskrit.items())
+			ab_str = '; '.join(pAdas_ab_result.imperfect_label_sanskrit.values())
+			cd_str = '; '.join(pAdas_cd_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: ?? {ab_str}; 3,4: {cd_str})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, one half imperfect, one length error)"]
 			Vrs.is_perfect = False
 			Vrs.diagnostic = {'ab': pAdas_ab_result, 'cd': pAdas_cd_result}
 			return pAdas_cd_result
 		elif pAdas_ab_result.imperfect() and pAdas_cd_result.length_error():
-			ab_str = '; '.join(f"{k}: {v}" for k, v in pAdas_ab_result.imperfect_label_sanskrit.items())
-			cd_str = '; '.join(f"{k}: {v}" for k, v in pAdas_cd_result.imperfect_label_sanskrit.items())
+			ab_str = '; '.join(pAdas_ab_result.imperfect_label_sanskrit.values())
+			cd_str = '; '.join(pAdas_cd_result.imperfect_label_sanskrit.values())
 			Vrs.meter_label = f"anuṣṭubh (1,2: {ab_str}; 3,4: ?? {cd_str})"
 			Vrs.identification_score = meter_scores["anuṣṭubh, full, one half imperfect, one length error)"]
 			Vrs.is_perfect = False
@@ -1972,3 +1998,68 @@ class MeterIdentifier(object):
 				bucket['_perfect_count'] = bucket.get('_perfect_count', 0) + 1
 
 		return V
+
+	def identify_meter_batch(self, rw_strs,
+		resplit_option=default_resplit_option,
+		resplit_keep_midpoint=default_resplit_keep_midpoint,
+		from_scheme=None):
+		"""
+		Parallel version of identify_meter() for a list of raw strings.
+
+		Spawns up to BATCH_MAX_WORKERS worker processes, one task per verse.
+		Returns a list of Verse objects in the same order as the input.
+		When _DEBUG_TIMING is on, merges per-verse timing dicts back into
+		the main process's _category_totals so flush_profiling_report() works.
+		Falls back to serial processing for small batches below BATCH_PARALLEL_THRESHOLD.
+		"""
+		if len(rw_strs) < BATCH_PARALLEL_THRESHOLD:
+			return [self.identify_meter(s, resplit_option=resplit_option,
+				resplit_keep_midpoint=resplit_keep_midpoint, from_scheme=from_scheme)
+				for s in rw_strs]
+
+		args = [(s, resplit_option, resplit_keep_midpoint, from_scheme, _DEBUG_TIMING) for s in rw_strs]
+		with ProcessPoolExecutor(max_workers=BATCH_MAX_WORKERS) as executor:
+			results = list(executor.map(_identify_meter_worker, args))
+
+		if _DEBUG_TIMING:
+			for V, verse_times, cat in results:
+				_section_totals['wiggle_count'] = _section_totals.get('wiggle_count', 0) + verse_times.pop('wiggle_count', 0)
+				bucket = _category_totals.setdefault(cat, {})
+				for k, v in verse_times.items():
+					bucket[k] = bucket.get(k, 0.0) + v
+				bucket['_count'] = bucket.get('_count', 0) + 1
+				if _verse_is_perfect(V):
+					bucket['_perfect_count'] = bucket.get('_perfect_count', 0) + 1
+			return [V for V, _, _ in results]
+
+		return results
+
+
+def _identify_meter_worker(args):
+	"""Module-level worker function (must be picklable). One verse per call."""
+	rw_str, resplit_option, resplit_keep_midpoint, from_scheme, debug_timing = args
+	if debug_timing:
+		import skrutable.utils as _utils
+		_utils._DEBUG_TIMING = True
+		import skrutable.meter_identification as _mi
+		_mi._DEBUG_TIMING = True
+	MI = MeterIdentifier()
+	all_keys = ('scan_clean', 'scan_translit', 'scan_syllabify', 'scan_weights', 'scan_morae_gana',
+		'anuzwuB', 'samavftta', 'upajAti', 'vizamavftta',
+		'ardhasamavftta_perfect', 'jAti', 'lev_samavftta', 'lev_ardha', 'lev_vizama', 'samavftta_etc')
+	if debug_timing:
+		pre = {k: _section_totals.get(k, 0.0) for k in all_keys}
+		pre_wiggle = _section_totals.get('wiggle_count', 0)
+	V = MI.identify_meter(
+		rw_str,
+		resplit_option=resplit_option,
+		resplit_keep_midpoint=resplit_keep_midpoint,
+		from_scheme=from_scheme,
+	)
+	if debug_timing:
+		verse_times = {k: _section_totals.get(k, 0.0) - pre[k] for k in all_keys}
+		verse_times['scan'] = sum(verse_times[k] for k in ('scan_clean', 'scan_translit', 'scan_syllabify', 'scan_weights', 'scan_morae_gana'))
+		verse_times['wiggle_count'] = _section_totals.get('wiggle_count', 0) - pre_wiggle
+		cat = _meter_label_to_category(V.meter_label)
+		return V, verse_times, cat
+	return V

{skrutable-2.6.1 → skrutable-2.6.3}/src/skrutable.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skrutable
-Version: 2.6.1
+Version: 2.6.3
 Summary: skrutable library for working with Sanskrit text
 Home-page: https://github.com/tylergneill/skrutable
 Author: Tyler Neill

skrutable-2.6.1/src/skrutable/__init__.py

@@ -1 +0,0 @@
-__version__ = "2.6.1"