merge w/ upstream

9 files changed, 560 insertions, 128 deletions

diff --git a/realtime/README.md b/realtime/README.md
index b37dddc8..e5290fc5 100644
--- a/realtime/README.md
+++ b/realtime/README.md
@@ -1 +1,17 @@
-More to come.
+cdec Realtime
+-------------
+
+Code by Michael Denkowski (http://www.cs.cmu.edu/~mdenkows/, mdenkows@cs.cmu.edu)
+
+```
+@misc{denkowski-proposal2013,
+    author       = {Michael Denkowski},
+    title        = {Machine Translation for Human Translators},
+    year         = {2013},
+    month        = {May},
+    day          = {30},
+    howpublished = {{Ph.D.} Thesis Proposal, Carnegie Mellon University}
+}
+```
+
+For a full tutorial, see http://www.cs.cmu.edu/~mdenkows/cdec-realtime.html
diff --git a/realtime/mkconfig.py b/realtime/mkconfig.py
index 32388978..f193c57a 100755
--- a/realtime/mkconfig.py
+++ b/realtime/mkconfig.py
@@ -4,9 +4,11 @@ import os
 import shutil
 import sys
 
+# Import first to make sure pycdec is on path
+import rt
+
 from cdec.configobj import ConfigObj
 
-import rt
 
 def main():
 
diff --git a/realtime/mkinput.py b/realtime/mkinput.py
new file mode 100755
index 00000000..a1b1256d
--- /dev/null
+++ b/realtime/mkinput.py
@@ -0,0 +1,18 @@
+#!/usr/bin/env python
+
+import itertools
+import sys
+
+def main():
+
+    if len(sys.argv[1:]) < 2:
+        sys.stderr.write('usage: {} test.src test.ref [ctx_name] >test.input\n'.format(sys.argv[0]))
+        sys.exit(2)
+
+    ctx_name = ' {}'.format(sys.argv[3]) if len(sys.argv[1:]) > 2 else ''
+    for (src, ref) in itertools.izip(open(sys.argv[1]), open(sys.argv[2])):
+        sys.stdout.write('TR{} ||| {}'.format(ctx_name, src))
+        sys.stdout.write('LEARN{} ||| {} ||| {}'.format(ctx_name, src.strip(), ref))
+
+if __name__ == '__main__':
+    main()
diff --git a/realtime/realtime.py b/realtime/realtime.py
index dff7e90c..ec15b59d 100755
--- a/realtime/realtime.py
+++ b/realtime/realtime.py
@@ -3,50 +3,109 @@
 import argparse
 import logging
 import sys
+import threading
+import time
 
 import rt
 
+ABOUT = '''Realtime adaptive translation with cdec (See README.md)
+
+Code by Michael Denkowski
+
+Citation:
+@misc{denkowski-proposal2013,
+    author       = {Michael Denkowski},
+    title        = {Machine Translation for Human Translators},
+    year         = {2013},
+    month        = {May},
+    day          = {30},
+    howpublished = {{Ph.D.} Thesis Proposal, Carnegie Mellon University}
+}
+
+'''
+
 class Parser(argparse.ArgumentParser):
 
     def error(self, message):
+        sys.stderr.write(ABOUT)
         self.print_help()
         sys.stderr.write('\n{}\n'.format(message))
         sys.exit(2)
 
+def handle_line(translator, line, output, ctx_name):
+    res = translator.command_line(line, ctx_name)
+    if res:
+        output.write('{}\n'.format(res))
+        output.flush()
+
+def test1(translator, input, output, ctx_name):
+    inp = open(input)
+    out = open(output, 'w')
+    for line in inp:
+        handle_line(translator, line.strip(), out, ctx_name)
+    out.close()
+
+def debug(translator, input):
+    # Test 1: multiple contexts
+    threads = []
+    for i in range(4):
+        t = threading.Thread(target=test1, args=(translator, input, '{}.out.{}'.format(input, i), str(i)))
+        threads.append(t)
+        t.start()
+        time.sleep(30)
+    # Test 2: flood
+    out = open('{}.out.flood'.format(input), 'w')
+    inp = open(input)
+    while True:
+        line = inp.readline()
+        if not line:
+            break
+        line = line.strip()
+        t = threading.Thread(target=handle_line, args=(translator, line.strip(), out, None))
+        threads.append(t)
+        t.start()
+        time.sleep(1) 
+    translator.drop_ctx(None)
+    # Join test threads
+    for t in threads:
+        t.join()
+
 def main():
 
-    parser = Parser(description='Real-time adaptive translation with cdec.')
-    parser.add_argument('-c', '--config', required=True, help='Config directory (see README.md)')
+    parser = Parser()
+    parser.add_argument('-c', '--config', required=True, help='Config directory')
+    parser.add_argument('-s', '--state', help='Load state file to default context (saved incremental data)')
     parser.add_argument('-n', '--normalize', help='Normalize text (tokenize, translate, detokenize)', action='store_true')
     parser.add_argument('-T', '--temp', help='Temp directory (default /tmp)', default='/tmp')
     parser.add_argument('-a', '--cache', help='Grammar cache size (default 5)', default='5')
     parser.add_argument('-v', '--verbose', help='Info to stderr', action='store_true')
+    parser.add_argument('-D', '--debug-test', help='Run debug tests on input file')
     args = parser.parse_args()
 
     if args.verbose:
         logging.basicConfig(level=logging.INFO)
 
-    rtd = rt.RealtimeDecoder(args.config, tmpdir=args.temp, cache_size=int(args.cache), norm=args.normalize)
+    with rt.RealtimeTranslator(args.config, tmpdir=args.temp, cache_size=int(args.cache), norm=args.normalize) as translator:
+
+        # Debugging
+        if args.debug_test:
+            debug(translator, args.debug_test)
+            return
+
+        # Load state if given
+        if args.state:
+            rtd.load_state(state)
 
-    try:
+        # Read lines and commands
         while True:
             line = sys.stdin.readline()
             if not line:
                 break
-            input = [f.strip() for f in line.split('|||')]
-            if len(input) == 1:
-                hyp = rtd.decode(input[0])
-                sys.stdout.write('{}\n'.format(hyp))
+            line = line.strip()
+            res = translator.command_line(line)
+            if res:
+                sys.stdout.write('{}\n'.format(res))
                 sys.stdout.flush()
-            elif len(input) == 2:
-                rtd.learn(*input)
-
-    # Clean exit on ctrl+c
-    except KeyboardInterrupt:
-        logging.info('Caught KeyboardInterrupt, exiting')
-
-    # Cleanup
-    rtd.close()
-
+     
 if __name__ == '__main__':
     main()
diff --git a/realtime/rt/__init__.py b/realtime/rt/__init__.py
index 738777f3..c76acc4d 100644
--- a/realtime/rt/__init__.py
+++ b/realtime/rt/__init__.py
@@ -1,3 +1,18 @@
+# Add pycdec to the Python path if user hasn't
+import os
+import sys
+try:
+    import cdec
+except ImportError as ie:
+    try:
+        pycdec = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))), 'python')
+        sys.path.append(pycdec)
+        import cdec
+    except:
+        sys.stderr.write('Error: cannot import pycdec.  Please check that cdec/python is built.\n')
+        raise ie
+
+# Regular init imports
 from rt import *
 import aligner
 import decoder
diff --git a/realtime/rt/aligner.py b/realtime/rt/aligner.py
index 3c6ea144..e1782496 100644
--- a/realtime/rt/aligner.py
+++ b/realtime/rt/aligner.py
@@ -2,14 +2,17 @@ import logging
 import os
 import sys
 import subprocess
+import threading
 
 import util
 
+logger = logging.getLogger('rt.aligner')
+
 class ForceAligner:
 
-    def __init__(self, fwd_params, fwd_err, rev_params, rev_err):
+    def __init__(self, fwd_params, fwd_err, rev_params, rev_err, heuristic='grow-diag-final-and'):
 
-        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
+        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
         fast_align = os.path.join(cdec_root, 'word-aligner', 'fast_align')
         atools = os.path.join(cdec_root, 'utils', 'atools')
 
@@ -18,21 +21,27 @@ class ForceAligner:
 
         fwd_cmd = [fast_align, '-i', '-', '-d', '-T', fwd_T, '-m', fwd_m, '-f', fwd_params]
         rev_cmd = [fast_align, '-i', '-', '-d', '-T', rev_T, '-m', rev_m, '-f', rev_params, '-r']
-        tools_cmd = [atools, '-i', '-', '-j', '-', '-c', 'grow-diag-final-and']
+        tools_cmd = [atools, '-i', '-', '-j', '-', '-c', heuristic]
 
-        logging.info('Executing: {}'.format(' '.join(fwd_cmd)))
+        logger.info('Executing: {}'.format(' '.join(fwd_cmd)))
         self.fwd_align = util.popen_io(fwd_cmd)
 
-        logging.info('Executing: {}'.format(' '.join(rev_cmd)))
+        logger.info('Executing: {}'.format(' '.join(rev_cmd)))
         self.rev_align = util.popen_io(rev_cmd)
 
-        logging.info('Executing: {}'.format(' '.join(tools_cmd)))
+        logger.info('Executing: {}'.format(' '.join(tools_cmd)))
         self.tools = util.popen_io(tools_cmd)
 
+        # Used to guarantee thread safety
+        self.lock = util.FIFOLock()
+
     def align(self, source, target):
+        '''Threadsafe, FIFO'''
         return self.align_formatted('{} ||| {}'.format(source, target))
 
     def align_formatted(self, line):
+        '''Threadsafe, FIFO'''
+        self.lock.acquire()
         self.fwd_align.stdin.write('{}\n'.format(line))
         self.rev_align.stdin.write('{}\n'.format(line))
         # f words ||| e words ||| links ||| score
@@ -40,12 +49,21 @@ class ForceAligner:
         rev_line = self.rev_align.stdout.readline().split('|||')[2].strip()
         self.tools.stdin.write('{}\n'.format(fwd_line))
         self.tools.stdin.write('{}\n'.format(rev_line))
-        return self.tools.stdout.readline().strip()
+        al_line = self.tools.stdout.readline().strip()
+        self.lock.release()
+        return al_line
  
-    def close(self):
+    def close(self, force=False):
+        if not force:
+            self.lock.acquire()
         self.fwd_align.stdin.close()
+        self.fwd_align.wait()
         self.rev_align.stdin.close()
+        self.rev_align.wait()
         self.tools.stdin.close()
+        self.tools.wait()
+        if not force:
+            self.lock.release()
 
     def read_err(self, err):
         (T, m) = ('', '')
diff --git a/realtime/rt/decoder.py b/realtime/rt/decoder.py
index 0a202fae..ed45c248 100644
--- a/realtime/rt/decoder.py
+++ b/realtime/rt/decoder.py
@@ -1,39 +1,80 @@
 import logging
 import os
 import subprocess
+import threading
 
 import util
 
+logger = logging.getLogger('rt.decoder')
+
 class Decoder:
 
-    def close(self):
+    def close(self, force=False):
+        if not force:
+            self.lock.acquire()
         self.decoder.stdin.close()
+        self.decoder.wait()
+        if not force:
+            self.lock.release()
 
-    def decode(self, sentence, grammar):
-        input = '<seg grammar="{g}">{s}</seg>\n'.format(s=sentence, g=grammar)
+    def decode(self, sentence, grammar=None):
+        '''Threadsafe, FIFO'''
+        self.lock.acquire()
+        input = '<seg grammar="{g}">{s}</seg>\n'.format(s=sentence, g=grammar) if grammar else '{}\n'.format(sentence)
         self.decoder.stdin.write(input)
-        return self.decoder.stdout.readline().strip()
+        hyp = self.decoder.stdout.readline().strip()
+        self.lock.release()
+        return hyp
 
 class CdecDecoder(Decoder):
-    
+
     def __init__(self, config, weights):
-        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
+        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
         decoder = os.path.join(cdec_root, 'decoder', 'cdec')
         decoder_cmd = [decoder, '-c', config, '-w', weights]
-        logging.info('Executing: {}'.format(' '.join(decoder_cmd)))
+        logger.info('Executing: {}'.format(' '.join(decoder_cmd)))
         self.decoder = util.popen_io(decoder_cmd)
+        self.lock = util.FIFOLock()
 
 class MIRADecoder(Decoder):
 
     def __init__(self, config, weights):
-        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
+        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
         mira = os.path.join(cdec_root, 'training', 'mira', 'kbest_cut_mira')
         #                                              optimizer=2 step=0.001    best=500,    k=500,       uniq, stream
         mira_cmd = [mira, '-c', config, '-w', weights, '-o', '2', '-C', '0.001', '-b', '500', '-k', '500', '-u', '-t']
-        logging.info('Executing: {}'.format(' '.join(mira_cmd)))
+        logger.info('Executing: {}'.format(' '.join(mira_cmd)))
         self.decoder = util.popen_io(mira_cmd)
+        self.lock = util.FIFOLock()
+
+    def get_weights(self):
+        '''Threadsafe, FIFO'''
+        self.lock.acquire()
+        self.decoder.stdin.write('WEIGHTS ||| WRITE\n')
+        weights = self.decoder.stdout.readline().strip()
+        self.lock.release()
+        return weights
+
+    def set_weights(self, w_line):
+        '''Threadsafe, FIFO'''
+        self.lock.acquire()
+        try:
+            # Check validity
+            for w_str in w_line.split():
+                (k, v) = w_str.split('=')
+                float(v)
+            self.decoder.stdin.write('WEIGHTS ||| {}\n'.format(w_line))
+            self.lock.release()
+        except:
+            self.lock.release()
+            raise Exception('Invalid weights line: {}'.format(w_line))
+
 
     def update(self, sentence, grammar, reference):
-        input = '<seg grammar="{g}">{s}</seg> ||| {r}\n'.format(s=sentence, g=grammar, r=reference)
+        '''Threadsafe, FIFO'''
+        self.lock.acquire()
+        input = 'LEARN ||| <seg grammar="{g}">{s}</seg> ||| {r}\n'.format(s=sentence, g=grammar, r=reference)
         self.decoder.stdin.write(input)
-        return self.decoder.stdout.readline().strip()
+        log = self.decoder.stdout.readline().strip()
+        self.lock.release()
+        return log
diff --git a/realtime/rt/rt.py b/realtime/rt/rt.py
index 892cc217..c0aec410 100644
--- a/realtime/rt/rt.py
+++ b/realtime/rt/rt.py
@@ -5,33 +5,88 @@ import collections
 import logging
 import os
 import shutil
-import sys
+import StringIO
 import subprocess
+import sys
 import tempfile
+import threading
 import time
 
-from cdec.configobj import ConfigObj
-import cdec.sa
-
+import cdec
 import aligner
 import decoder
 import util
 
+# Dummy input token that is unlikely to appear in normalized data (but no fatal errors if it does)
+LIKELY_OOV = '(OOV)'
+
+logger = logging.getLogger('rt')
+
 class RealtimeDecoder:
+    '''Do not use directly unless you know what you're doing.  Use RealtimeTranslator.'''
+
+    def __init__(self, configdir, tmpdir):
+
+        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+        self.tmp = tmpdir
+        os.mkdir(self.tmp)
+
+        # HPYPLM reference stream
+        ref_fifo_file = os.path.join(self.tmp, 'ref.fifo')
+        os.mkfifo(ref_fifo_file)
+        self.ref_fifo = open(ref_fifo_file, 'w+')
+        # Start with empty line (do not learn prior to first input)
+        self.ref_fifo.write('\n')
+        self.ref_fifo.flush()
+
+        # Decoder
+        decoder_config = [[f.strip() for f in line.split('=')] for line in open(os.path.join(configdir, 'cdec.ini'))]
+        util.cdec_ini_for_realtime(decoder_config, os.path.abspath(configdir), ref_fifo_file)
+        decoder_config_file = os.path.join(self.tmp, 'cdec.ini')
+        with open(decoder_config_file, 'w') as output:
+            for (k, v) in decoder_config:
+                output.write('{}={}\n'.format(k, v))
+        decoder_weights = os.path.join(configdir, 'weights.final')
+        self.decoder = decoder.MIRADecoder(decoder_config_file, decoder_weights)
+
+    def close(self, force=False):
+        logger.info('Closing decoder and removing {}'.format(self.tmp))
+        self.decoder.close(force)
+        self.ref_fifo.close()
+        shutil.rmtree(self.tmp)
+
+class RealtimeTranslator:
+    '''Main entry point into API: serves translations to any number of concurrent users'''
 
     def __init__(self, configdir, tmpdir='/tmp', cache_size=5, norm=False):
 
-        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
+        # name -> (method, set of possible nargs)
+        self.COMMANDS = {
+                'TR': (self.translate, set((1,))),
+                'LEARN': (self.learn, set((2,))),
+                'SAVE': (self.save_state, set((0, 1))),
+                'LOAD': (self.load_state, set((0, 1))),
+                'DROP': (self.drop_ctx, set((0,))),
+                'LIST': (self.list_ctx, set((0,))),
+                }
+        
+        cdec_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+        ### Single instance for all contexts
 
+        self.config = configdir
         # Temporary work dir
         self.tmp = tempfile.mkdtemp(dir=tmpdir, prefix='realtime.')
-        logging.info('Using temp dir {}'.format(self.tmp))
+        logger.info('Using temp dir {}'.format(self.tmp))
 
         # Normalization
         self.norm = norm
         if self.norm:
             self.tokenizer = util.popen_io([os.path.join(cdec_root, 'corpus', 'tokenize-anything.sh'), '-u'])
+            self.tokenizer_lock = util.FIFOLock()
             self.detokenizer = util.popen_io([os.path.join(cdec_root, 'corpus', 'untok.pl')])
+            self.detokenizer_lock = util.FIFOLock()
 
         # Word aligner
         fwd_params = os.path.join(configdir, 'a.fwd_params')
@@ -41,112 +96,302 @@ class RealtimeDecoder:
         self.aligner = aligner.ForceAligner(fwd_params, fwd_err, rev_params, rev_err)
 
         # Grammar extractor
-        sa_config = ConfigObj(os.path.join(configdir, 'sa.ini'), unrepr=True)
+        sa_config = cdec.configobj.ConfigObj(os.path.join(configdir, 'sa.ini'), unrepr=True)
         sa_config.filename = os.path.join(self.tmp, 'sa.ini')
         util.sa_ini_for_realtime(sa_config, os.path.abspath(configdir))
         sa_config.write()
         self.extractor = cdec.sa.GrammarExtractor(sa_config.filename, online=True)
-        self.grammar_files = collections.deque()
-        self.grammar_dict = {}
         self.cache_size = cache_size
 
-        # HPYPLM reference stream
-        ref_fifo_file = os.path.join(self.tmp, 'ref.fifo')
-        os.mkfifo(ref_fifo_file)
-        self.ref_fifo = open(ref_fifo_file, 'w+')
-        # Start with empty line (do not learn prior to first input)
-        self.ref_fifo.write('\n')
-        self.ref_fifo.flush()
+        ### One instance per context
 
-        # Decoder
-        decoder_config = [[f.strip() for f in line.split('=')] for line in open(os.path.join(configdir, 'cdec.ini'))]
-        util.cdec_ini_for_realtime(decoder_config, os.path.abspath(configdir), ref_fifo_file)
-        decoder_config_file = os.path.join(self.tmp, 'cdec.ini')
-        with open(decoder_config_file, 'w') as output:
-            for (k, v) in decoder_config:
-                output.write('{}={}\n'.format(k, v))
-        decoder_weights = os.path.join(configdir, 'weights.final')
-        self.decoder = decoder.MIRADecoder(decoder_config_file, decoder_weights)
+        self.ctx_names = set()
+        # All context-dependent operations are atomic
+        self.ctx_locks = collections.defaultdict(util.FIFOLock)
+        # ctx -> list of (source, target, alignment)
+        self.ctx_data = {}
 
-    def close(self):
-        logging.info('Closing processes')
-        self.aligner.close()
-        self.decoder.close()
-        self.ref_fifo.close()
+        # Grammar extractor is not threadsafe
+        self.extractor_lock = util.FIFOLock()
+        # ctx -> deque of file
+        self.grammar_files = {}
+        # ctx -> dict of {sentence: file}
+        self.grammar_dict = {}
+
+        self.decoders = {}
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, ex_type, ex_value, ex_traceback):
+        # Force shutdown on exception
+        self.close(ex_type is not None)
+
+    def close(self, force=False):
+        '''Cleanup'''
+        if force:
+            logger.info('Forced shutdown: stopping immediately')
+        for ctx_name in list(self.ctx_names):
+            self.drop_ctx(ctx_name, force)
+        logger.info('Closing processes')
+        self.aligner.close(force)
         if self.norm:
+            if not force:
+                self.tokenizer_lock.acquire()
+                self.detokenizer_lock.acquire()
             self.tokenizer.stdin.close()
+            self.tokenizer.wait()
             self.detokenizer.stdin.close()
-        logging.info('Deleting {}'.format(self.tmp))
+            self.detokenizer.wait()
+            if not force:
+                self.tokenizer_lock.release()
+                self.detokenizer_lock.release()
+        logger.info('Deleting {}'.format(self.tmp))
         shutil.rmtree(self.tmp)
 
-    def grammar(self, sentence):
-        grammar_file = self.grammar_dict.get(sentence, None)
+    def lazy_ctx(self, ctx_name):
+        '''Initialize a context (inc starting a new decoder) if needed.
+        NOT threadsafe, acquire ctx_name lock before calling.'''
+        if ctx_name in self.ctx_names:
+            return
+        logger.info('({}) New context'.format(ctx_name))
+        self.ctx_names.add(ctx_name)
+        self.ctx_data[ctx_name] = []
+        self.grammar_files[ctx_name] = collections.deque()
+        self.grammar_dict[ctx_name] = {}
+        tmpdir = os.path.join(self.tmp, 'decoder.{}'.format(ctx_name))
+        self.decoders[ctx_name] = RealtimeDecoder(self.config, tmpdir)
+
+    def drop_ctx(self, ctx_name=None, force=False):
+        '''Delete a context (inc stopping the decoder)
+        Threadsafe and FIFO unless forced.'''
+        lock = self.ctx_locks[ctx_name]
+        if not force:
+            lock.acquire()
+        if ctx_name not in self.ctx_names:
+            logger.info('({}) No context found, no action taken'.format(ctx_name))
+            if not force:
+                lock.release()
+            return
+        logger.info('({}) Dropping context'.format(ctx_name))
+        self.ctx_names.remove(ctx_name)
+        self.ctx_data.pop(ctx_name)
+        self.extractor.drop_ctx(ctx_name)
+        self.grammar_files.pop(ctx_name)
+        self.grammar_dict.pop(ctx_name)
+        self.decoders.pop(ctx_name).close(force)
+        self.ctx_locks.pop(ctx_name)
+        if not force:
+            lock.release()
+
+    def list_ctx(self, ctx_name=None):
+        '''Return a string of active contexts'''
+        return 'ctx_name ||| {}'.format(' '.join(sorted(str(ctx_name) for ctx_name in self.ctx_names)))
+
+    def grammar(self, sentence, ctx_name=None):
+        '''Extract a sentence-level grammar on demand (or return cached)
+        Threadsafe wrt extractor but NOT decoder.  Acquire ctx_name lock
+        before calling.'''
+        self.extractor_lock.acquire()
+        self.lazy_ctx(ctx_name)
+        grammar_dict = self.grammar_dict[ctx_name]
+        grammar_file = grammar_dict.get(sentence, None)
         # Cache hit
         if grammar_file:
-            logging.info('Grammar cache hit')
+            logger.info('({}) Grammar cache hit: {}'.format(ctx_name, grammar_file))
+            self.extractor_lock.release()
             return grammar_file
         # Extract and cache
-        grammar_file = tempfile.mkstemp(dir=self.tmp, prefix='grammar.')[1]
+        (fid, grammar_file) = tempfile.mkstemp(dir=self.decoders[ctx_name].tmp, prefix='grammar.')
+        os.close(fid)
         with open(grammar_file, 'w') as output:
-            for rule in self.extractor.grammar(sentence):
+            for rule in self.extractor.grammar(sentence, ctx_name):
                 output.write('{}\n'.format(str(rule)))
-        if len(self.grammar_files) == self.cache_size:
-            rm_sent = self.grammar_files.popleft()
+        grammar_files = self.grammar_files[ctx_name]
+        if len(grammar_files) == self.cache_size:
+            rm_sent = grammar_files.popleft()
             # If not already removed by learn method
-            if rm_sent in self.grammar_dict:
-                rm_grammar = self.grammar_dict.pop(rm_sent)
+            if rm_sent in grammar_dict:
+                rm_grammar = grammar_dict.pop(rm_sent)
                 os.remove(rm_grammar)
-        self.grammar_files.append(sentence)
-        self.grammar_dict[sentence] = grammar_file
+        grammar_files.append(sentence)
+        grammar_dict[sentence] = grammar_file
+        self.extractor_lock.release()
         return grammar_file
         
-    def decode(self, sentence):
+    def translate(self, sentence, ctx_name=None):
+        '''Decode a sentence (inc extracting a grammar if needed)
+        Threadsafe, FIFO'''
+        lock = self.ctx_locks[ctx_name]
+        lock.acquire()
+        self.lazy_ctx(ctx_name)
         # Empty in, empty out
         if sentence.strip() == '':
+            lock.release()
             return ''
         if self.norm:
             sentence = self.tokenize(sentence)
-            logging.info('Normalized input: {}'.format(sentence))
-        grammar_file = self.grammar(sentence)
+            logger.info('({}) Normalized input: {}'.format(ctx_name, sentence))
+        grammar_file = self.grammar(sentence, ctx_name)
+        decoder = self.decoders[ctx_name]
         start_time = time.time()
-        hyp = self.decoder.decode(sentence, grammar_file)
+        hyp = decoder.decoder.decode(sentence, grammar_file)
         stop_time = time.time()
-        logging.info('Translation time: {} seconds'.format(stop_time - start_time))
+        logger.info('({}) Translation time: {} seconds'.format(ctx_name, stop_time - start_time))
         # Empty reference: HPYPLM does not learn prior to next translation
-        self.ref_fifo.write('\n')
-        self.ref_fifo.flush()
+        decoder.ref_fifo.write('\n')
+        decoder.ref_fifo.flush()
         if self.norm:
-            logging.info('Normalized translation: {}'.format(hyp))
+            logger.info('({}) Normalized translation: {}'.format(ctx_name, hyp))
             hyp = self.detokenize(hyp)
+        lock.release()
         return hyp
 
     def tokenize(self, line):
+        self.tokenizer_lock.acquire()
         self.tokenizer.stdin.write('{}\n'.format(line))
-        return self.tokenizer.stdout.readline().strip()
+        tok_line = self.tokenizer.stdout.readline().strip()
+        self.tokenizer_lock.release()
+        return tok_line
 
     def detokenize(self, line):
+        self.detokenizer_lock.acquire()
         self.detokenizer.stdin.write('{}\n'.format(line))
-        return self.detokenizer.stdout.readline().strip()
+        detok_line = self.detokenizer.stdout.readline().strip()
+        self.detokenizer_lock.release()
+        return detok_line
 
-    def learn(self, source, target):
+    def command_line(self, line, ctx_name=None):
+        # COMMAND [ctx_name] ||| arg1 [||| arg2 ...]
+        args = [f.strip() for f in line.split('|||')]
+        if args[-1] == '':
+            args = args[:-1]
+        if len(args) > 0:
+            cmd_name = args[0].split()
+            # ctx_name provided
+            if len(cmd_name) == 2:
+                (cmd_name, ctx_name) = cmd_name
+            # ctx_name default/passed
+            else:
+                cmd_name = cmd_name[0]
+            (command, nargs) = self.COMMANDS.get(cmd_name, (None, None))
+            if command and len(args[1:]) in nargs:
+                logger.info('({}) {} ||| {}'.format(ctx_name, cmd_name, ' ||| '.join(args[1:])))
+                return command(*args[1:], ctx_name=ctx_name)
+        logger.info('ERROR: command: {}'.format(' ||| '.join(args)))
+        
+    def learn(self, source, target, ctx_name=None):
+        '''Learn from training instance (inc extracting grammar if needed)
+        Threadsafe, FIFO'''
+        lock = self.ctx_locks[ctx_name]
+        lock.acquire()
+        self.lazy_ctx(ctx_name)
         if '' in (source.strip(), target.strip()):
-            logging.info('Error empty source or target: {} ||| {}'.format(source, target))
+            logger.info('({}) ERROR: empty source or target: {} ||| {}'.format(ctx_name, source, target))
+            lock.release()
             return
         if self.norm:
             source = self.tokenize(source)
             target = self.tokenize(target)
+        # Align instance
+        alignment = self.aligner.align(source, target)
+        grammar_file = self.grammar(source, ctx_name)
         # MIRA update before adding data to grammar extractor
-        grammar_file = self.grammar(source)
-        mira_log = self.decoder.update(source, grammar_file, target)
-        logging.info('MIRA: {}'.format(mira_log))
+        decoder = self.decoders[ctx_name]
+        mira_log = decoder.decoder.update(source, grammar_file, target)
+        logger.info('({}) MIRA HBF: {}'.format(ctx_name, mira_log))
+        # Add to HPYPLM by writing to fifo (read on next translation)
+        logger.info('({}) Adding to HPYPLM: {}'.format(ctx_name, target))
+        decoder.ref_fifo.write('{}\n'.format(target))
+        decoder.ref_fifo.flush()
+        # Store incremental data for save/load
+        self.ctx_data[ctx_name].append((source, target, alignment))
         # Add aligned sentence pair to grammar extractor
-        alignment = self.aligner.align(source, target)
-        logging.info('Adding instance: {} ||| {} ||| {}'.format(source, target, alignment))
-        self.extractor.add_instance(source, target, alignment)
+        logger.info('({}) Adding to bitext: {} ||| {} ||| {}'.format(ctx_name, source, target, alignment))
+        self.extractor.add_instance(source, target, alignment, ctx_name)
         # Clear (old) cached grammar
-        rm_grammar = self.grammar_dict.pop(source)
+        rm_grammar = self.grammar_dict[ctx_name].pop(source)
         os.remove(rm_grammar)
-        # Add to HPYPLM by writing to fifo (read on next translation)
-        logging.info('Adding to HPYPLM: {}'.format(target))
-        self.ref_fifo.write('{}\n'.format(target))
-        self.ref_fifo.flush()
+        lock.release()
+
+    def save_state(self, file_or_stringio=None, ctx_name=None):
+        '''Write state (several lines terminated by EOF line) to file, buffer, or stdout'''
+        lock = self.ctx_locks[ctx_name]
+        lock.acquire()
+        self.lazy_ctx(ctx_name)
+        ctx_data = self.ctx_data[ctx_name]
+        # Filename, StringIO or None (stdout)
+        if file_or_stringio:
+            if isinstance(file_or_stringio, StringIO.StringIO):
+                out = file_or_stringio
+            else:
+                out = open(file_or_stringio, 'w')
+        else:
+            out = sys.stdout
+        logger.info('({}) Saving state with {} sentences'.format(ctx_name, len(ctx_data)))
+        out.write('{}\n'.format(self.decoders[ctx_name].decoder.get_weights()))
+        for (source, target, alignment) in ctx_data:
+            out.write('{} ||| {} ||| {}\n'.format(source, target, alignment))
+        out.write('EOF\n')
+        # Close if file
+        if file_or_stringio and not isinstance(file_or_stringio, StringIO.StringIO):
+            out.close()
+        lock.release()
+
+    def load_state(self, file_or_stringio=None, ctx_name=None):
+        '''Load state (several lines terminated by EOF line) from file, buffer, or stdin.
+        Restarts context on any error.'''
+        lock = self.ctx_locks[ctx_name]
+        lock.acquire() 
+        self.lazy_ctx(ctx_name)
+        ctx_data = self.ctx_data[ctx_name]
+        decoder = self.decoders[ctx_name]
+        # Filename, StringIO, or None (stdin)
+        if file_or_stringio:
+            if isinstance(file_or_stringio, StringIO.StringIO):
+                input = file_or_stringio
+            else:
+                input = open(file_or_stringio)
+        else:
+            input = sys.stdin
+        # Non-initial load error
+        if ctx_data:
+            logger.info('({}) ERROR: Incremental data has already been added to context'.format(ctx_name))
+            logger.info('  State can only be loaded to a new context.')
+            lock.release()
+            return
+        # Many things can go wrong if bad state data is given
+        try:
+            # MIRA weights
+            line = input.readline().strip()
+            # Throws exception if bad line
+            decoder.decoder.set_weights(line)
+            logger.info('({}) Loading state...'.format(ctx_name))
+            start_time = time.time()
+            # Lines source ||| target ||| alignment
+            while True:
+                line = input.readline()
+                if not line:
+                    raise Exception('End of file before EOF line')
+                line = line.strip()
+                if line == 'EOF':
+                    break
+                (source, target, alignment) = line.split(' ||| ')
+                ctx_data.append((source, target, alignment))
+                # Extractor
+                self.extractor.add_instance(source, target, alignment, ctx_name)
+                # HPYPLM
+                hyp = decoder.decoder.decode(LIKELY_OOV)
+                decoder.ref_fifo.write('{}\n'.format(target))
+                decoder.ref_fifo.flush()
+            stop_time = time.time()
+            logger.info('({}) Loaded state with {} sentences in {} seconds'.format(ctx_name, len(ctx_data), stop_time - start_time))
+            lock.release()
+        # Recover from bad load attempt by restarting context.
+        # Guaranteed not to cause data loss since only a new context can load state.
+        except:
+            logger.info('({}) ERROR: could not load state, restarting context'.format(ctx_name))
+            # ctx_name is already owned and needs to be restarted before other blocking threads use
+            self.drop_ctx(ctx_name, force=True)
+            self.lazy_ctx(ctx_name)
+            lock.release()
diff --git a/realtime/rt/util.py b/realtime/rt/util.py
index 10e94909..52767dac 100644
--- a/realtime/rt/util.py
+++ b/realtime/rt/util.py
@@ -1,4 +1,5 @@
 import os
+import Queue
 import subprocess
 import sys
 import threading
@@ -13,29 +14,58 @@ SA_INI_FILES = set((
     'precompute_file',
     ))
 
+class FIFOLock:
+    '''Lock that preserves FIFO order of blocking threads'''
+
+    def __init__(self):
+        self.q = Queue.Queue()
+        self.i = 0
+        self.lock = threading.Lock()
+
+    def acquire(self):
+        self.lock.acquire()
+        self.i += 1
+        if self.i > 1:
+            event = threading.Event()
+            self.q.put(event)
+            self.lock.release()
+            event.wait()
+            return
+        self.lock.release()
+
+    def release(self):
+        self.lock.acquire()
+        self.i -= 1
+        if self.i > 0:
+            self.q.get().set()
+        self.lock.release()
+
 def cdec_ini_for_config(config):
-    cdec_ini_handle(config, os.path.basename, hpyplm_rm_ref)
+    # This is a list of (k, v), not a ConfigObj or dict
+    for i in range(len(config)):
+        if config[i][0] == 'feature_function':
+            if config[i][1].startswith('KLanguageModel'):
+                f = config[i][1].split()
+                f[-1] = 'mono.klm'
+                config[i][1] = ' '.join(f)
+            elif config[i][1].startswith('External'):
+                config[i][1] = 'External libcdec_ff_hpyplm.so corpus.hpyplm'
 
 def cdec_ini_for_realtime(config, path, ref_fifo):
-    cdec_ini_handle(config, lambda x: os.path.join(path, x), lambda x: hpyplm_add_ref(x, ref_fifo))
-
-def cdec_ini_handle(config, path_fn, hpyplm_fn):
     # This is a list of (k, v), not a ConfigObj or dict
     for i in range(len(config)):
         if config[i][0] == 'feature_function':
             if config[i][1].startswith('KLanguageModel'):
                 f = config[i][1].split()
-                f[-1] = path_fn(f[-1])
+                f[-1] = os.path.join(path, f[-1])
                 config[i][1] = ' '.join(f)
             elif config[i][1].startswith('External'):
                 f = config[i][1].split()
-                if f[1].endswith('libcdec_ff_hpyplm.so'):
-                    # Modify paths
-                    for j in range(1, len(f)):
-                        if not f[j].startswith('-'):
-                            f[j] = path_fn(f[j])
-                    # Modify hpyplm args
-                    hpyplm_fn(f)
+                f[1] =  os.path.join(path, f[1])
+                f[2] =  os.path.join(path, f[2])
+                f.append('-r')
+                f.append(ref_fifo)
+                f.append('-t')
                 config[i][1] = ' '.join(f)
 
 def consume_stream(stream):
@@ -44,18 +74,6 @@ def consume_stream(stream):
             pass
     threading.Thread(target=consume, args=(stream,)).start()
 
-def hpyplm_add_ref(f, ref):
-    f.append('-r')
-    f.append(ref)
-    f.append('-t')
-
-def hpyplm_rm_ref(f):
-    for i in range(1, len(f)):
-        if f[i] == '-r':
-            f.pop(i)
-            f.pop(i)
-            return
-
 def popen_io(cmd):
     p = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
     consume_stream(p.stderr)