skrutable 2.6.0__tar.gz → 2.6.2__tar.gz

{skrutable-2.6.0 → skrutable-2.6.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skrutable
-Version: 2.6.0
+Version: 2.6.2
 Summary: skrutable library for working with Sanskrit text
 Home-page: https://github.com/tylergneill/skrutable
 Author: Tyler Neill

skrutable-2.6.2/src/skrutable/__init__.py

@@ -0,0 +1 @@
+__version__ = "2.6.2"

{skrutable-2.6.0 → skrutable-2.6.2}/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')
@@ -1972,3 +1985,68 @@ 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-2.6.0 → skrutable-2.6.2}/src/skrutable/phonemes.py

@@ -33,6 +33,10 @@ For transliteration, 'consonant' means 'needs virāma if non-vowel follows' (no
 SLP_consonants_for_scansion = SLP_consonants
 """For scansion, 'consonant' means 'contributes to heaviness of previous vowel' (yes M H)"""
 
+SLP_vowels_set = set(SLP_vowels)
+SLP_long_vowels_set = set(SLP_long_vowels)
+SLP_consonants_for_scansion_set = set(SLP_consonants_for_scansion)
+
 DEV_consonants = ['क', 'ख', 'ग', 'घ', 'ङ','च', 'छ', 'ज', 'झ', 'ञ',
 'ट', 'ठ', 'ड', 'ढ', 'ण','त', 'थ', 'द', 'ध', 'न','प', 'फ', 'ब', 'भ', 'म',
 'य', 'र', 'ल', 'व','श', 'ष', 'स', 'ह']
@@ -111,3 +115,5 @@ to_add = [' ', '\t', '\n']
 for k in character_set.keys():
 	for c in to_add:
 		character_set[k].append(c)
+
+character_set_lookup = {k: set(v) for k, v in character_set.items()}

{skrutable-2.6.0 → skrutable-2.6.2}/src/skrutable/scansion.py

@@ -6,6 +6,9 @@ from skrutable.config import load_config_dict_from_json_file
 from skrutable.utils import timed
 import re
 
+_re_ws_around_newline = re.compile(r'[ \t]*\n[ \t]*')
+_re_multi_newline = re.compile(r'\n+')
+
 # load config variables
 config = load_config_dict_from_json_file()
 scansion_syllable_separator = config["scansion_syllable_separator"] # e.g. " "
@@ -160,8 +163,8 @@ class Scanner(object):
 		# filter out disallowed characters (numbers, irrelevant punctuation, etc.)
 		# pāda separator chars are preserved so they can be converted to \n below
 		pAda_sep_chars = set(c for sep in additional_pAda_separators for c in sep)
-		for c in list(set(cntnts)):
-			if c not in phonemes.character_set[scheme_in] and c not in pAda_sep_chars:
+		for c in set(cntnts):
+			if c not in phonemes.character_set_lookup[scheme_in] and c not in pAda_sep_chars:
 				cntnts = cntnts.replace(c, '')
 
 		# replace all pāda separator strings with newline
@@ -169,8 +172,8 @@ class Scanner(object):
 			cntnts = cntnts.replace(sep, '\n')
 
 		# strip horizontal whitespace around newlines, dedupe, strip leading/trailing
-		cntnts = re.sub(r'[ \t]*\n[ \t]*', '\n', cntnts)
-		cntnts = re.sub(r'\n+', '\n', cntnts)
+		cntnts = _re_ws_around_newline.sub('\n', cntnts)
+		cntnts = _re_multi_newline.sub('\n', cntnts)
 		cntnts = cntnts.strip()
 
 		return cntnts
@@ -207,15 +210,15 @@ class Scanner(object):
 			# place scansion_syllable_separator after vowels
 			for letter in line:
 
-				# exception: do treat M and H as explicit syllable coda
-				if letter in ['M', 'H']:
-					if line_syllables[-1] == scansion_syllable_separator:
+				if letter in ('M', 'H'):
+					# M and H are explicit syllable codas: strip any trailing separator, append, re-add separator
+					if line_syllables and line_syllables[-1] == scansion_syllable_separator:
 						line_syllables = line_syllables[:-1]
-
-				line_syllables += letter
-
-				if letter in phonemes.SLP_vowels + ['M', 'H']:
-					line_syllables += scansion_syllable_separator
+					line_syllables += letter + scansion_syllable_separator
+				elif letter in phonemes.SLP_vowels_set:
+					line_syllables += letter + scansion_syllable_separator
+				else:
+					line_syllables += letter
 
 			# e.g. 'ya.dA.ya.dA.hi.Da.rma.sya.glA.ni.rBa.va.ti.BA.ra.ta.'
 			# BUT e.g. 'a.Byu.tTA.na.ma.Da.rma.sya.ta.dA.tmA.na.Msf.jA.mya.ha.m'
@@ -256,8 +259,6 @@ class Scanner(object):
 
 		for line in text_lines:
 
-			line_weights = ''
-
 			syllables = line.split(scansion_syllable_separator)
 
 			try:
@@ -265,34 +266,37 @@ class Scanner(object):
 					syllables.pop(-1) # in case of final separator(s)
 			except IndexError: pass
 
+			line_weights_chars = []
+
 			for n, syllable in enumerate(syllables):
 
 				if (
 					# heavy by nature
-					syllable[-1] in phonemes.SLP_long_vowels + ['M', 'H']
+					syllable[-1] in phonemes.SLP_long_vowels_set or syllable[-1] in ('M', 'H')
 
 					or
 
 					# heavy by position:
 					# consonant closes syllable or next syllable begins with a cluster
-					syllable[-1] in (phonemes.SLP_consonants_for_scansion)
+					syllable[-1] in phonemes.SLP_consonants_for_scansion_set
 					or
 					n <= (len(syllables)-2)
 					and len(syllables[n+1]) > 1
-					and syllables[n+1][0] in (phonemes.SLP_consonants_for_scansion)
-					and syllables[n+1][1] in (phonemes.SLP_consonants_for_scansion)
+					and syllables[n+1][0] in phonemes.SLP_consonants_for_scansion_set
+					and syllables[n+1][1] in phonemes.SLP_consonants_for_scansion_set
+
 
 					):
 
-					line_weights += 'g'
-					# line_weights += 'g_'
+					line_weights_chars.append('g')
+					# line_weights_chars.append('g_')
 					# insofar as two 'l's can equal one 'g', could use this alternative for better visual alignment
 
 				else:
 
-					line_weights += 'l'
+					line_weights_chars.append('l')
 
-			weights_by_line.append(line_weights)
+			weights_by_line.append(''.join(line_weights_chars))
 
 		syllable_weights = '\n'.join(weights_by_line) # restore newlines
 		return syllable_weights
@@ -325,21 +329,16 @@ class Scanner(object):
 		Returns string of 'gaRa'-trisyllable abbreviation, e.g. 'nml'.
 		"""
 
-		for c in list(set(syl_wts)):
-			if c not in ['l','g']:
+		for c in set(syl_wts):
+			if c not in {'l', 'g'}:
 				return None
 
-		weights_of_curr_gaRa = ''
-		overall_abbreviation = ''
-
-		for single_weight in syl_wts:
-			weights_of_curr_gaRa += single_weight
-			if len(weights_of_curr_gaRa) == 3:
-				overall_abbreviation += meter_patterns.gaRas_by_weights[weights_of_curr_gaRa]
-				weights_of_curr_gaRa = ''
-
+		n = len(syl_wts) // 3 * 3
+		overall_abbreviation = ''.join(
+			meter_patterns.gaRas_by_weights[syl_wts[i:i+3]] for i in range(0, n, 3)
+		)
 		# leftover lights and heavies (l/g)
-		overall_abbreviation += weights_of_curr_gaRa
+		overall_abbreviation += syl_wts[n:]
 
 		return overall_abbreviation
 

{skrutable-2.6.0 → skrutable-2.6.2}/src/skrutable.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skrutable
-Version: 2.6.0
+Version: 2.6.2
 Summary: skrutable library for working with Sanskrit text
 Home-page: https://github.com/tylergneill/skrutable
 Author: Tyler Neill

skrutable-2.6.0/src/skrutable/__init__.py

@@ -1 +0,0 @@
-__version__ = "2.6.0"