merge paul's extractor code

3 files changed, 28 insertions, 4 deletions

diff --git a/python/pkg/cdec/sa/compile.py b/python/pkg/cdec/sa/compile.py
index ce249c0f..d4cd8387 100644
--- a/python/pkg/cdec/sa/compile.py
+++ b/python/pkg/cdec/sa/compile.py
@@ -4,6 +4,7 @@ import os
 import logging
 import cdec.configobj
 import cdec.sa
+from cdec.sa._sa import monitor_cpu
 import sys
 
 MAX_PHRASE_LENGTH = 4
@@ -21,6 +22,7 @@ def precompute(f_sa, max_len, max_nt, max_size, min_gap, rank1, rank2, tight_phr
     return precomp
 
 def main():
+    preprocess_start_time = monitor_cpu()
     sys.setrecursionlimit(sys.getrecursionlimit() * 100)
 
     logging.basicConfig(level=logging.INFO)
@@ -73,31 +75,46 @@ def main():
     a_bin = os.path.join(args.output, 'a.bin')
     lex_bin = os.path.join(args.output, 'lex.bin')
 
+    start_time = monitor_cpu()
     logger.info('Compiling source suffix array')
     if args.bitext:
         f_sa = cdec.sa.SuffixArray(from_text=args.bitext, side='source')
     else:
         f_sa = cdec.sa.SuffixArray(from_text=args.source)
     f_sa.write_binary(f_sa_bin)
+    stop_time = monitor_cpu()
+    logger.info('Compiling source suffix array took %f seconds', stop_time - start_time)
 
+    start_time = monitor_cpu()
     logger.info('Compiling target data array')
     if args.bitext:
         e = cdec.sa.DataArray(from_text=args.bitext, side='target')
     else:
         e = cdec.sa.DataArray(from_text=args.target)
     e.write_binary(e_bin)
+    stop_time = monitor_cpu()
+    logger.info('Compiling target data array took %f seconds', stop_time - start_time)
 
+    start_time = monitor_cpu()
     logger.info('Precomputing frequent phrases')
     precompute(f_sa, *params).write_binary(precomp_bin)
+    stop_time = monitor_cpu()
+    logger.info('Compiling precomputations took %f seconds', stop_time - start_time)
 
+    start_time = monitor_cpu()
     logger.info('Compiling alignment')
     a = cdec.sa.Alignment(from_text=args.alignment)
     a.write_binary(a_bin)
+    stop_time = monitor_cpu()
+    logger.info('Compiling alignment took %f seonds', stop_time - start_time)
 
+    start_time = monitor_cpu()
     logger.info('Compiling bilexical dictionary')
     lex = cdec.sa.BiLex(from_data=True, alignment=a, earray=e, fsarray=f_sa)
     lex.write_binary(lex_bin)
-    
+    stop_time = monitor_cpu()
+    logger.info('Compiling bilexical dictionary took %f seconds', stop_time - start_time)
+
     # Write configuration
     config = cdec.configobj.ConfigObj(args.config, unrepr=True)
     config['f_sa_file'] = os.path.abspath(f_sa_bin)
@@ -108,6 +125,8 @@ def main():
     for name, value in zip(param_names, params):
         config[name] = value
     config.write()
+    preprocess_stop_time = monitor_cpu()
+    logger.info('Overall preprocessing step took %f seconds', preprocess_stop_time - preprocess_start_time)
 
 if __name__ == '__main__':
     main()
diff --git a/python/pkg/cdec/sa/extract.py b/python/pkg/cdec/sa/extract.py
index bf39d080..b6502c52 100644
--- a/python/pkg/cdec/sa/extract.py
+++ b/python/pkg/cdec/sa/extract.py
@@ -8,6 +8,7 @@ import logging
 import signal
 import multiprocessing as mp
 import cdec.sa
+from cdec.sa._sa import monitor_cpu
 
 extractor, prefix = None, None
 online, compress = False, False
@@ -88,9 +89,10 @@ def main():
             sys.stderr.write('Error: feature definition file <{}>'
                     ' should be a python module\n'.format(featdef))
             sys.exit(1)
-    
+
     online = args.online
 
+    start_time = monitor_cpu()
     if args.jobs > 1:
         logging.info('Starting %d workers; chunk size: %d', args.jobs, args.chunksize)
         pool = mp.Pool(args.jobs, make_extractor, (args,))
@@ -104,5 +106,8 @@ def main():
         for output in map(extract, enumerate(sys.stdin)):
             print(output)
 
+    stop_time = monitor_cpu()
+    logging.info("Overall extraction step took %f seconds", stop_time - start_time)
+
 if __name__ == '__main__':
     main()
diff --git a/python/pkg/cdec/sa/features.py b/python/pkg/cdec/sa/features.py
index 46412cd5..c8fc1cca 100644
--- a/python/pkg/cdec/sa/features.py
+++ b/python/pkg/cdec/sa/features.py
@@ -105,7 +105,7 @@ def IsSingletonF(ctx):
         count = ctx.fcount
     else:
         count = ctx.fcount + ctx.online.fcount  
-    return (count == 1)
+    return math.fabs(count - 1) < 1e-6
 
 def IsSingletonFE(ctx):
     if not ctx.online:
@@ -139,4 +139,4 @@ def IsSupportedOnline(ctx): # Occurs in online data?
     if ctx.online:
         return (ctx.online.paircount > 0.01)
     else:
-        return False
\ No newline at end of file
+        return False