Merge pull request #172 from structuredllm/json_quote

Fix ByteFSM for dead state

Author: shubhamugare

syncode/mask_store/byte_fsm.py

@@ -50,7 +50,7 @@ def _build_byte_fsm(self, regex_fsm):
         self.transitions = {}
 
         # Create a mapping from byte values to category numbers
-        self.byte_to_category = {}
+        self.byte_to_category = {}        
 
         # Extract the mapping from the regex FSM's alphabet and build our byte-level alphabet
         for char, category in regex_fsm.alphabet.items():
@@ -84,8 +84,8 @@ def _build_byte_fsm(self, regex_fsm):
         # Copy the transitions from the regex FSM to our byte FSM
         for state, category_transitions in regex_fsm.map.items():
             for category, target in category_transitions.items():
-                self.transitions[state][category] = target
-        
+                self.transitions[state][category] = target          
+
         # Handle multi-byte Unicode characters separately
         # This is needed because a multi-byte character might need special handling
         for char, category in regex_fsm.alphabet.items():
@@ -95,30 +95,42 @@ def _build_byte_fsm(self, regex_fsm):
             char_bytes = char.encode('utf-8')
             if len(char_bytes) <= 1:
                 continue
-                
+            
+            # Add an explicit dead state for invalid transitions
+            dead_state = f"DEAD"
+            if dead_state not in self.transitions:
+                self.transitions[dead_state] = {}
+
             # For multi-byte characters, we need to add special transitions
             # Make a copy of states to avoid modifying the dictionary during iteration
             states_to_process = list(self.transitions.keys())
             for state in states_to_process:
                 if category in self.transitions[state]:
                     target = self.transitions[state][category]
+                else:
+                    target = dead_state
 
-                    # Create intermediate states for the multi-byte character
-                    current = state
-                    for i, byte in enumerate(char_bytes):
-                        if byte not in self.alphabet:
-                            # Add the byte to the alphabet with a new category
-                            byte_category = f"{byte}_{i}"
-                            self.byte_to_category[byte] = byte_category
+                # Create intermediate states for the multi-byte character
+                current = state
+                for i, byte in enumerate(char_bytes):
+                    if byte not in self.alphabet:
+                        # Add the byte to the alphabet with a new category
+                        byte_category = f"{byte}_{i}"
+                        self.byte_to_category[byte] = byte_category
 
-                        if i < len(char_bytes) - 1:
+                    if i < len(char_bytes) - 1:
+                        if byte_category not in self.transitions[current]:
+                            # Create a new state for this byte
                             next_state = f"{current}_{byte}_{i}_{char}"
                             if next_state not in self.transitions:
                                 self.transitions[next_state] = {}
                             self.transitions[current][byte_category] = next_state
                             current = next_state
                         else:
-                            self.transitions[current][byte_category] = target
+                            # Transition already exists
+                            current = self.transitions[current][byte_category]
+                    else:
+                        self.transitions[current][byte_category] = target
 
     @lru_cache(maxsize=100000)
     def _get_category(self, byte_val: int) -> Any:

syncode/mask_store/fsm_set.py

@@ -1,6 +1,6 @@
 import time
 import interegular
-from typing import Any, Optional, Tuple, Iterable, Dict
+from typing import Any, Optional, Tuple, Iterable, Dict, Union
 from syncode.mask_store.byte_fsm import ByteFSM
 import logging
 logger = logging.getLogger(__name__)
@@ -21,11 +21,20 @@ def __hash__(self):
         return self._hash
 
     @staticmethod
-    def det_hash(terminal: str, state_id: int):
+    def det_hash(terminal: str, state_id: Union[str, int]):
         h = 0
         for char in terminal:
             h = (h * 31 + ord(char)) & 0xFFFFFFFF
-        h = (h * 31 + state_id) & 0xFFFFFFFF
+        
+        # Handle state_id based on its type
+        if isinstance(state_id, str):
+            # If state_id is a string, hash each character
+            for char in state_id:
+                h = (h * 31 + ord(char)) & 0xFFFFFFFF
+        else:
+            # If state_id is an integer, hash it directly
+            h = (h * 31 + state_id) & 0xFFFFFFFF
+        
         return h
 
     def __repr__(self):

tests/mask_store/test_byte_fsm.py

@@ -120,6 +120,9 @@ def test_consume_prefix(self):
                 ("[email protected]", (False, None)),
                 ("user@", (True, b"")),  # Live state
                 ("invalid", (True, b""))  # Live state for [a-z]+
+            ]),
+            ('"[^"”“]+"', [
+                ('\"key”', (False, None)),
             ])
         ]
 

tests/mask_store/test_mask_store_go.py

@@ -0,0 +1,107 @@
+import sys
+import os
+import time
+import unittest
+sys.path.append(os.path.dirname(os.path.realpath(__file__)) + '/../..')
+import syncode.common as common
+from syncode.parsers.incremental_parser import ParseResult
+from syncode.parse_result import AcceptSequence, RemainderState
+from syncode.mask_store.mask_store import MaskStore
+from syncode.parsers.grammars.grammar import Grammar
+from tests.test_utils import CustomAssertMixin
+
+
+class TestMaskGo(unittest.TestCase, CustomAssertMixin):
+    def setUp(self):
+        model = 'Qwen/Qwen2.5-1.5B-Instruct'
+        tokenizer = common.load_tokenizer(model)
+        self.mask_store = MaskStore.init_mask_store(grammar=Grammar('go'), tokenizer=tokenizer, use_cache=False, mode='grammar_mask')
+        return super().setUp()
+
+    def test_mask(self):
+        r = ParseResult({AcceptSequence(['DECIMAL_LIT', 'PLUS'])}, b'1', RemainderState.MAYBE_COMPLETE)
+        self.mask_store.get_accept_mask(r, get_list=True)
+        result_list = self.mask_store.get_accept_mask(r, get_list=True)
+        for token in [' +', ' +=', ' ++']:
+            self.assertInWithLimit(token, result_list, f"{token} not found in result list")
+
+    def test_mask2(self):
+        r = ParseResult({AcceptSequence(['EOS'])}, b'\n // 1.', RemainderState.MAYBE_COMPLETE)
+        result_list = self.mask_store.get_accept_mask(r, get_list=True)
+        self.assertTrue(len(result_list) > 32000, "Result list is smaller than expected")
+
+    def test_mask3(self):
+        r = ParseResult({AcceptSequence(['__ANON_14'])}, b'', RemainderState.COMPLETE)
+        result_list = self.mask_store.get_accept_mask(r, get_list=True)
+        # Uncomment the following line if you want to assert presence of specific tokens
+        self.assertInWithLimit(":=", result_list, ":= not found in result list")
+
+    def test_mask4(self):
+        r = ParseResult({AcceptSequence(['__IGNORE_0'])}, b'', RemainderState.COMPLETE)
+        self.assertInWithLimit("\t", self.mask_store.get_accept_mask(r, get_list=True), "Tab character not found in result list")
+
+    def test_mask5(self):
+        r = ParseResult({AcceptSequence(['LBRACE', '__IGNORE_0'])}, b'{', RemainderState.MAYBE_COMPLETE)
+        self.assertInWithLimit("\t", self.mask_store.get_accept_mask(r, get_list=True), "Tab character not found in result list")
+
+    def test_mask6(self):
+        r = ParseResult({AcceptSequence(['NAME'])}, b'for', RemainderState.MAYBE_COMPLETE)
+        self.assertInWithLimit(" {", self.mask_store.get_accept_mask(r, get_list=True), "Opening brace not found in result list")
+
+
+class TestMaskJSON(unittest.TestCase, CustomAssertMixin):
+    def setUp(self):
+        model = 'google/gemma-2-2b-it'
+        tokenizer = common.load_tokenizer(model)
+
+        custom_json_grammar = f"""
+        ?start: start_value
+        ?start_value: object
+        | array
+
+        ?value: object
+        | array
+        | EMPTY_STRING
+        | NONEMPTY_STRING
+        | SIGNED_NUMBER      -> number
+        | "true"             -> true
+        | "false"            -> false
+        | "null"             -> null
+
+        array  : "[" [value ("," value)*] "]"
+        object : "{" [pair ("," pair)*] "}"
+        pair   : NONEMPTY_STRING ":" value
+
+        NONEMPTY_STRING: /\"[^"”“]+\"/
+        EMPTY_STRING: /\"\"/
+
+        DIGIT: "0".."9"
+        HEXDIGIT: "a".."f"|"A".."F"|DIGIT
+        INT: DIGIT+
+        SIGNED_INT: ["+"|"-"] INT
+        DECIMAL: INT "." INT? | "." INT
+
+
+        _EXP: ("e"|"E") SIGNED_INT
+        FLOAT: INT _EXP | DECIMAL _EXP?
+        NUMBER: FLOAT | INT
+        SIGNED_NUMBER: ["+"|"-"] NUMBER
+        WS: /[ \t\f\r\n]/+
+
+        %ignore WS
+        """
+        self.mask_store = MaskStore.init_mask_store(grammar=Grammar(custom_json_grammar), tokenizer=tokenizer, use_cache=False, mode='grammar_mask')
+        return super().setUp()
+    
+    def test_mask(self):
+        r = ParseResult({AcceptSequence(['NONEMPTY_STRING'])}, b'"key', RemainderState.INCOMPLETE)
+        result_list = self.mask_store.get_accept_mask(r, get_list=True)
+        self.assertInWithLimit('"', result_list, '" not found in result list')
+        self.assertNotIn('”', result_list)
+        self.assertNotIn('“', result_list)        
+
+
+if __name__ == '__main__':
+    # Run JSON tests
+    suite = unittest.TestLoader().loadTestsFromTestCase(TestMaskJSON)
+    unittest.TextTestRunner().run(suite)

tests/mask_store/test_mask_store_misc.py

@@ -19,7 +19,7 @@ class TestDFAMaskLlama(unittest.TestCase, CustomAssertMixin):
     mask_store = MaskStore.init_mask_store(
         grammar=Grammar('python'), 
         tokenizer=tokenizer, 
-        use_cache=True, 
+        use_cache=False, 
         indent=True, 
         mode="grammar_strict"
         )

tests/mask_store/test_mask_store_python.py

@@ -25,4 +25,13 @@ def test_json_parser2(self):
         r = inc_parser.get_acceptable_next_terminals(partial_code)
         assert r.remainder == ''
         assert r.remainder_state == RemainderState.COMPLETE
-        
+    
+    def test_json_parser3(self):
+        # Tests when the last incomplete word is unparsed
+        inc_parser.reset()
+        partial_code = '{\n  "key'
+        r = inc_parser.get_acceptable_next_terminals(partial_code)
+        assert AcceptSequence(['NONEMPTY_STRING']) in r.accept_sequences
+
+if __name__ == '__main__':
+    unittest.main()