Fix/14 caching issues using dictionary (#15)

fix: caching point issue (#14)

Author: vimt

README.md

@@ -10,14 +10,16 @@
 
 # MaxMind-DB-Writer-python
 
-Make `mmdb` format ip library file which can be read by [`maxmind` official language reader](https://dev.maxmind.com/geoip/geoip2/downloadable/)
+Make `mmdb` format ip library file which can be read by [
+`maxmind` official language reader](https://dev.maxmind.com/geoip/geoip2/downloadable/)
 
-~~[The official perl writer](https://github.com/maxmind/MaxMind-DB-Writer-perl) was written in perl, 
-which was difficult to customize. 
+~~[The official perl writer](https://github.com/maxmind/MaxMind-DB-Writer-perl) was written in perl,
+which was difficult to customize.
 So I implemented the `MaxmindDB format` ip library in python language.~~
 
-MaxMind has now released an official Go version of the MMDB writer. 
-If you prefer using Go, you can check out the official Go implementation [mmdbwriter](https://github.com/maxmind/mmdbwriter). 
+MaxMind has now released an official Go version of the MMDB writer.
+If you prefer using Go, you can check out the official Go
+implementation [mmdbwriter](https://github.com/maxmind/mmdbwriter).
 This project still provides a Python alternative for those who need it.
 
 ## Install
@@ -27,30 +29,35 @@ pip install -U mmdb_writer
 ```
 
 ## Usage
+
 ```python
 from netaddr import IPSet
 
 from mmdb_writer import MMDBWriter
+
 writer = MMDBWriter()
 
 writer.insert_network(IPSet(['1.1.0.0/24', '1.1.1.0/24']), {'country': 'COUNTRY', 'isp': 'ISP'})
 writer.to_db_file('test.mmdb')
 
 import maxminddb
+
 m = maxminddb.open_database('test.mmdb')
 r = m.get('1.1.1.1')
 assert r == {'country': 'COUNTRY', 'isp': 'ISP'}
 ```
 
 ## Examples
+
 see [csv_to_mmdb.py](./examples/csv_to_mmdb.py)
 Here is a professional and clear translation of the README.md section from Chinese into English:
 
 ## Using the Java Client
 
-### TLDR
+If you are using the Java client, you need to be careful to set the `int_type` parameter so that Java correctly
+recognizes the integer type in the MMDB file.
 
-When generating an MMDB file for use with the Java client, you must specify the `int_type`:
+Example:
 
 ```python
 from mmdb_writer import MMDBWriter
@@ -65,15 +72,15 @@ Alternatively, you can explicitly specify data types using the [Type Enforcement
 In Java, when deserializing to a structure, the numeric types will use the original MMDB numeric types. The specific
 conversion relationships are as follows:
 
-| mmdb type    | java type  |
-|--------------|------------|
-| float (15)   | Float      |
-| double (3)   | Double     |
-| int32 (8)    | Integer    |
-| uint16 (5)   | Integer    |
-| uint32 (6)   | Long       |
-| uint64 (9)   | BigInteger |
-| uint128 (10) | BigInteger |
+| mmdb type | java type  |
+|-----------|------------|
+| float     | Float      |
+| double    | Double     |
+| int32     | Integer    |
+| uint16    | Integer    |
+| uint32    | Long       |
+| uint64    | BigInteger |
+| uint128   | BigInteger |
 
 When using the Python writer to generate an MMDB file, by default, it converts integers to the corresponding MMDB type
 based on the size of the `int`. For instance, `int(1)` would convert to `uint16`, and `int(2**16+1)` would convert
@@ -97,7 +104,17 @@ MMDB file. The behaviors for different `int_type` settings are:
 | u64            | Stores all integer types as `uint64`.                                                                                                                                                                                                                         |
 | u128           | Stores all integer types as `uint128`.                                                                                                                                                                                                                        |
 
+If you want to use different int types for different scenarios, you can use type wrapping:
+
+```python
+from mmdb_writer import MMDBWriter, MmdbI32, MmdbF32
+
+writer = MMDBWriter()
+# the value of field "i32" will be stored as int32 type
+writer.insert_network(IPSet(["1.0.0.0/24"]), {"i32": MmdbI32(128), "f32": MmdbF32(1.22)})
+```
+
+## Reference:
 
-## Reference: 
 - [MaxmindDB format](http://maxmind.github.io/MaxMind-DB/)
 - [geoip-mmdb](https://github.com/i-rinat/geoip-mmdb)

mmdb_writer.py

@@ -378,11 +378,12 @@ def encode_meta(self, meta):
             res += self.encode(v, meta_type.get(k))
         return res
 
-    def encode(self, value, type_id=None):
+    def encode(self, value, type_id=None, return_offset=False):
         if self.cache:
             cache_key = self._freeze(value)
             try:
-                return self.data_cache[cache_key]
+                offset = self.data_cache[cache_key]
+                return offset if return_offset else self._encode_pointer(offset)
             except KeyError:
                 pass
 
@@ -399,18 +400,11 @@ def encode(self, value, type_id=None):
         res = encoder(value)
 
         if self.cache:
-            # add to cache
-            if type_id == 1:
-                self.data_list.append(res)
-                self.data_pointer += len(res)
-                return res
-            else:
-                self.data_list.append(res)
-                pointer_position = self.data_pointer
-                self.data_pointer += len(res)
-                pointer = self.encode(pointer_position, 1)
-                self.data_cache[cache_key] = pointer
-                return pointer
+            self.data_list.append(res)
+            offset = self.data_pointer
+            self.data_pointer += len(res)
+            self.data_cache[cache_key] = offset
+            return offset if return_offset else self._encode_pointer(offset)
         return res
 
 
@@ -484,8 +478,8 @@ def _enumerate_nodes(self, node):
         elif type(node) is SearchTreeLeaf:
             node_id = id(node)
             if node_id not in self._leaf_offset:
-                res = self.encoder.encode(node.value)
-                self._leaf_offset[node_id] = self._data_pointer - len(res)
+                offset = self.encoder.encode(node.value, return_offset=True)
+                self._leaf_offset[node_id] = offset + 16
         else:  # == None
             return