json-repair 0.3.0__py3-none-any.whl → 0.4.1__py3-none-any.whl

json_repair/json_repair.py

@@ -52,13 +52,17 @@ class JSONParser:
         # <array> starts with '['
         elif char == "[":
             return self.parse_array()
+        # there can be an edge case in which a key is empty and at the end of an object
+        # like "key": }. We return an empty string here to close the object properly
+        elif char == "}" and self.context == "object_value":
+            return ""
         # <string> starts with '"'
         elif char == '"':
             return self.parse_string()
         # <number> starts with [0-9] or minus
         elif char.isdigit() or char == "-":
             return self.parse_number()
-        # <boolean> could (T)rue or (F)alse or (N)ull
+        # <boolean> could be (T)rue or (F)alse or (N)ull
         elif char == "t" or char == "f" or char == "n":
             return self.parse_boolean_or_null()
         # This might be a <string> that is missing the starting '"'
@@ -67,6 +71,7 @@ class JSONParser:
         # Ignore whitespaces outside of strings
         elif char.isspace():
             self.index += 1
+            self.skip_whitespaces_at()
             return self.parse_json()
         # If everything else fails, then we give up and return an exception
         else:
@@ -87,7 +92,7 @@ class JSONParser:
             # <member> ::= <string> ': ' <json>
 
             # Skip filler whitespaces
-            self.trim()
+            self.skip_whitespaces_at()
 
             # Sometimes LLMs do weird things, if we find a ":" so early, we'll change it to "," and move on
             if self.get_char_at() == ":":
@@ -100,6 +105,7 @@ class JSONParser:
             self.context = "object_key"
 
             # <member> starts with a <string>
+            self.skip_whitespaces_at()
             key = self.parse_string()
             while key == "":
                 key = self.parse_string()
@@ -124,7 +130,7 @@ class JSONParser:
                 self.index += 1
 
             # Remove trailing spaces
-            self.trim()
+            self.skip_whitespaces_at()
 
         # Especially at the end of an LLM generated json you might miss the last "}"
         if self.get_char_at() and self.get_char_at() != "}":
@@ -172,8 +178,6 @@ class JSONParser:
         # Flag to manage corner cases related to missing starting quote
         fixed_quotes = False
         # i.e. { name: "John" }
-        # Remove any trailing space
-        self.trim()
         if self.get_char_at() != '"':
             self.insert_char_at('"')
             fixed_quotes = True
@@ -211,14 +215,10 @@ class JSONParser:
             and self.get_char_at().isspace()
         ):
             # skip whitespaces
-            self.trim()
+            self.skip_whitespaces_at()
             # This string is invalid if there's no valid termination afterwards
 
-            if (
-                self.get_char_at() != ":"
-                or self.get_char_at() != ","
-                or self.get_char_at() != "}"
-            ):
+            if self.get_char_at() not in [":", ","]:
                 return ""
 
         end = self.index
@@ -232,11 +232,13 @@ class JSONParser:
 
     def parse_number(self) -> Union[float, int]:
         # <number> is a valid real number expressed in one of a number of given formats
-        number_pattern = r"-?\d+(\.\d+)?([eE][+-]?\d+)?"
-        match = re.match(number_pattern, self.json_str[self.index :])
-        if match:
-            number_str = match.group()
-            self.index += len(number_str)
+        number_str = ""
+        char = self.get_char_at()
+        while char and (char.isdigit() or char in "-.eE"):
+            number_str += char
+            self.index += 1
+            char = self.get_char_at()
+        if number_str:
             if "." in number_str or "e" in number_str or "E" in number_str:
                 return float(number_str)
             else:
@@ -264,23 +266,34 @@ class JSONParser:
         self.json_str = self.json_str[: self.index] + char + self.json_str[self.index :]
         self.index += 1
 
-    def get_char_at(self, idx=0) -> Union[str, bool]:
+    def get_char_at(self) -> Union[str, bool]:
         # Why not use something simpler? Because we might be out of bounds and doing this check all the time is annoying
-        idx = self.index + idx
-        return self.json_str[idx] if idx < len(self.json_str) else False
+        try:
+            return self.json_str[self.index]
+        except IndexError:
+            return False
 
-    def remove_char_at(self, idx=0) -> None:
-        idx += 1
-        self.json_str = self.json_str[: self.index] + self.json_str[self.index + idx :]
+    def remove_char_at(self) -> None:
+        self.json_str = self.json_str[: self.index] + self.json_str[self.index + 1 :]
 
-    def trim(self) -> None:
+    def skip_whitespaces_at(self) -> None:
         # Remove trailing spaces
-        while self.get_char_at() and self.get_char_at().isspace():
+        # I'd rather not do this BUT this method is called so many times that it makes sense to expand get_char_at
+        # At least this is what the profiler said and I believe in our lord and savior the profiler
+        try:
+            char = self.json_str[self.index]
+        except IndexError:
+            return
+        while char and char.isspace():
             self.index += 1
+            try:
+                char = self.json_str[self.index]
+            except IndexError:
+                return
 
 
 def repair_json(
-    json_str: str, return_objects: bool = False
+    json_str: str, return_objects: bool = False, skip_json_loads: bool = False
 ) -> Union[Dict[str, Any], List[Any], str, float, int, bool, None]:
     """
     Given a json formatted string, it will try to decode it and, if it fails, it will try to fix it.
@@ -290,11 +303,14 @@ def repair_json(
     json_str = re.sub(r"^\s+", "", json_str)
     json_str = re.sub(r"\s+$", "", json_str)
     json_str = re.sub(r"/\*.*?\*/", "", json_str)
-    try:
-        parsed_json = json.loads(json_str)
-    except Exception:
-        parser = JSONParser(json_str)
+    parser = JSONParser(json_str)
+    if skip_json_loads:
         parsed_json = parser.parse()
+    else:
+        try:
+            parsed_json = json.loads(json_str)
+        except json.JSONDecodeError:
+            parsed_json = parser.parse()
     # It's useful to return the actual object instead of the json string, it allows this lib to be a replacement of the json library
     if return_objects:
         return parsed_json

{json_repair-0.3.0.dist-info → json_repair-0.4.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: json-repair
-Version: 0.3.0
+Version: 0.4.1
 Summary: A package to repair broken json strings
 Author-email: Stefano Baccianella <4247706+mangiucugna@users.noreply.github.com>
 License: MIT License
@@ -59,7 +59,7 @@ You can look how I used it by checking out this demo: https://huggingface.co/spa
         # Not even this library could fix this JSON
 
 You can use this library to completely replace `json.loads()`:
-    
+
     import json_repair
     try:
         decoded_object = json_repair.loads(json_string)
@@ -74,6 +74,21 @@ or just
     except Exception:
         # Manage Exception
 
+## Performance
+If you find this library too slow because is using `json.loads()` you can skip that by passing `skip_json_loads=True` to `repair_json`. Like:
+
+    from json_repair import repair_json
+    try:
+        good_json_string = repair_json(bad_json_string, skip_json_loads=True)
+    except Exception:
+        # Not even this library could fix this JSON
+
+I made a choice of not using any fast json library to avoid having any external dependency, so that anybody can use it regardless of their stack.
+
+Some rules of thumb to use:
+- Setting `return_objects=True` will always be faster because the parser returns an object already and it doesn't have serialize that object to JSON
+- `skip_json_loads` is faster only if you 100% know that the string is not a valid JSON
+
 # How it works
 This module will parse the JSON file following the BNF definition:
 

json_repair-0.4.1.dist-info/RECORD

@@ -0,0 +1,7 @@
+json_repair/__init__.py,sha256=p9mZnte8Bg18NcxqgJ7vopH2gQv_XbZ0dRnk686QuRE,92
+json_repair/json_repair.py,sha256=Qxy8eQpkm9e1qVUDvhYOuID7flLHIzAsr5cB1NWb3Y4,12974
+json_repair-0.4.1.dist-info/LICENSE,sha256=wrjQo8MhNrNCicXtMe3MHmS-fx8AmQk1ue8AQwiiFV8,1076
+json_repair-0.4.1.dist-info/METADATA,sha256=MXc1lBLV9wr-DNTqqkzuH8W-VYiDNY9UVsvsHJGqMw4,6155
+json_repair-0.4.1.dist-info/WHEEL,sha256=Xo9-1PvkuimrydujYJAjF7pCkriuXBpUPEjma1nZyJ0,92
+json_repair-0.4.1.dist-info/top_level.txt,sha256=7-VZwZN2CgB_n0NlSLk-rEUFh8ug21lESbsblOYuZqw,12
+json_repair-0.4.1.dist-info/RECORD,,

json_repair-0.3.0.dist-info/RECORD

@@ -1,7 +0,0 @@
-json_repair/__init__.py,sha256=p9mZnte8Bg18NcxqgJ7vopH2gQv_XbZ0dRnk686QuRE,92
-json_repair/json_repair.py,sha256=QdgEnpE4bDqy9nCR9Gw61MB6igfI4K0YFOrBI_dh5mc,12224
-json_repair-0.3.0.dist-info/LICENSE,sha256=wrjQo8MhNrNCicXtMe3MHmS-fx8AmQk1ue8AQwiiFV8,1076
-json_repair-0.3.0.dist-info/METADATA,sha256=EECNyro1Jtknlrtf3vqhl3StFJ_8drVdpTOodvZl7Us,5386
-json_repair-0.3.0.dist-info/WHEEL,sha256=Xo9-1PvkuimrydujYJAjF7pCkriuXBpUPEjma1nZyJ0,92
-json_repair-0.3.0.dist-info/top_level.txt,sha256=7-VZwZN2CgB_n0NlSLk-rEUFh8ug21lESbsblOYuZqw,12
-json_repair-0.3.0.dist-info/RECORD,,