diff --git a/ciftools/binary/decoder.py b/ciftools/binary/decoder.py
index 906c719..50f8766 100644
--- a/ciftools/binary/decoder.py
+++ b/ciftools/binary/decoder.py
@@ -1,3 +1,6 @@
+from numba import jit
+from numba.typed import Dict
+from numba.core import types
 import numpy as np
 from ciftools.binary.data_types import DataType
 from ciftools.binary.encoded_data import EncodedCIFData
@@ -47,12 +50,12 @@ def _decode_delta(data: np.ndarray, encoding: DeltaEncoding) -> np.ndarray:
     return np.cumsum(result, out=result)
 
 
-# TODO: JIT
-def _decode_integer_packing_signed(data: np.ndarray, encoding: IntegerPackingEncoding) -> np.ndarray:
-    upper_limit = 0x7F if encoding["byteCount"] == 1 else 0x7FFF
+@jit(nopython=True)
+def _decode_integer_packing_signed(data: np.ndarray, encoding_byte_count, encoding_src_size) -> np.ndarray:
+    upper_limit = 0x7F if encoding_byte_count == 1 else 0x7FFF
     lower_limit = -upper_limit - 1
     n = len(data)
-    output = np.zeros(encoding["srcSize"], dtype="i4")
+    output = np.zeros(encoding_src_size, dtype="i4")
     i = 0
     j = 0
     while i < n:
@@ -69,21 +72,29 @@ def _decode_integer_packing_signed(data: np.ndarray, encoding: IntegerPackingEnc
     return output
 
 
-# TODO: JIT
-def _decode_integer_packing_unsigned(data: np.ndarray, encoding: IntegerPackingEncoding) -> np.ndarray:
-    upper_limit = 0xFF if encoding["byteCount"] == 1 else 0xFFFF
+@jit(nopython=True)
+def _decode_integer_packing_unsigned(data: np.ndarray, encoding_byte_count, encoding_src_size) -> np.ndarray:
+    upper_limit = 0xFF if encoding_byte_count == 1 else 0xFFFF
     n = len(data)
-    output = np.zeros(encoding["srcSize"], dtype="i4")
+    output = np.zeros(encoding_src_size, dtype="i4")
     i = 0
     j = 0
+    print(data)
     while i < n:
         value = 0
+        # 255
         t = data[i]
+        # yes
         while t == upper_limit:
+            # 255
             value += t
+            # 1
             i += 1
+            # 145
             t = data[i]
+        # 255 + 145 = 400
         value += t
+        # 400
         output[j] = value
         i += 1
         j += 1
@@ -93,10 +104,24 @@ def _decode_integer_packing_unsigned(data: np.ndarray, encoding: IntegerPackingE
 def _decode_integer_packing(data: np.ndarray, encoding: IntegerPackingEncoding) -> np.ndarray:
     if len(data) == encoding["srcSize"]:
         return data
+
+    # d = dict()
+    # for k, v in encoding.items():
+    #     d[k] = v
+
+    
+    # d1 = Dict.empty(
+    #     key_type=types.py2_string_type,
+    #     value_type=types.int64,
+    # )
+
+    # for k, v in encoding.items():
+    #     d1[k] = v
+
     if encoding["isUnsigned"]:
-        return _decode_integer_packing_unsigned(data, encoding)
+        return _decode_integer_packing_unsigned(data, encoding["byteCount"], encoding["srcSize"])
     else:
-        return _decode_integer_packing_signed(data, encoding)
+        return _decode_integer_packing_signed(data, encoding["byteCount"], encoding["srcSize"])
 
 
 def _decode_string_array(data: np.ndarray, encoding: StringArrayEncoding) -> np.ndarray:
diff --git a/ciftools/binary/encoder.py b/ciftools/binary/encoder.py
index 30c067c..35669a5 100644
--- a/ciftools/binary/encoder.py
+++ b/ciftools/binary/encoder.py
@@ -1,3 +1,4 @@
+import itertools
 import math
 import sys
 from typing import Any, Dict, List, Protocol, Tuple, Union
@@ -304,34 +305,96 @@ def encode(self, data: np.ndarray) -> EncodedCIFData:
 
 class StringArray(BinaryCIFEncoder):
     def encode(self, data: Union[np.ndarray, List[str]]) -> EncodedCIFData:
+        # strings: List[str] = []
+        # offsets = [0]
+        # indices = np.empty(len(data), dtype="<i4")
+
+        # _pack_strings(
+        #     data,
+        #     indices,
+        #     strings,
+        #     offsets,
+        # )
+
+        # string_data = "".join(strings)
+
+        string_data, indices, offsets = _pack_strings(data)
+
+        encoded_offsets = _OFFSET_ENCODER.encode(np.array(offsets, dtype="<i4"))
+        encoded_data = _DATA_ENCODER.encode(indices)
+
+        encoding: StringArrayEncoding = {
+            "dataEncoding": encoded_data["encoding"],
+            "kind": EncodingEnun.StringArray,
+            "stringData": string_data,
+            "offsetEncoding": encoded_offsets["encoding"],
+            "offsets": encoded_offsets["data"],  # type: ignore
+        }
+
+        return EncodedCIFData(data=encoded_data["data"], encoding=[encoding])
+
+    def encode_not_optimized(self, data: Union[np.ndarray, List[str]]) -> EncodedCIFData:
         strings: List[str] = []
         offsets = [0]
         indices = np.empty(len(data), dtype="<i4")
 
-        _pack_strings(
+        _pack_strings_old(
             data,
             indices,
             strings,
             offsets,
         )
 
+        string_data = "".join(strings)
+
         encoded_offsets = _OFFSET_ENCODER.encode(np.array(offsets, dtype="<i4"))
         encoded_data = _DATA_ENCODER.encode(indices)
 
         encoding: StringArrayEncoding = {
             "dataEncoding": encoded_data["encoding"],
             "kind": EncodingEnun.StringArray,
-            "stringData": "".join(strings),
+            "stringData": string_data,
             "offsetEncoding": encoded_offsets["encoding"],
             "offsets": encoded_offsets["data"],  # type: ignore
         }
 
         return EncodedCIFData(data=encoded_data["data"], encoding=[encoding])
 
+# @jit(nopython=True)
+def _pack_strings(data: List[str]) -> Tuple[str, np.ndarray, np.ndarray]:
+    # strings = set(data)
+    # # mapping of strings to numbers
+    # str_map = {s: i for i, s in enumerate(strings)}
+    # string_data = "".join(strings)
+
+    # indices = np.array([str_map[s] for s in data], dtype='<i4')
+    # offset_data = np.empty(len(strings) + 1, dtype='<i4')
+    # offset_data[0] = 0
+    # # np.cumsum([len(s) for s in strings], out=offset_data[1:])
+    # # NOTE: this also result in errors
+    # # print(f'Strings: {strings}')
+    # # NOTE: this not
+    # # print('Strings', strings)
+    # # temp = np.array([len(s) for s in strings])
+    # # print(temp)
+    
+    # offset_data[1:] = np.cumsum(np.array([len(s) for s in strings]))
+    # return string_data, indices, 
+    strings = set(data)
+    str_map = {s: i for i, s in enumerate(strings)}
+    string_data = "".join(strings)
+
+    indices = np.array([str_map[s] for s in data], dtype='<i4')
+    offset_data = np.empty(len(strings) + 1, dtype='<i4')
+    offset_data[0] = 0
+    np.cumsum([len(s) for s in strings], out=offset_data[1:])
+
+    return string_data, indices, offset_data
+
 
 # TODO: benchmark if JIT helps here
-@jit(nopython=False, forceobj=True)
-def _pack_strings(data: List[str], indices: np.ndarray, strings: List[str], offsets: List[int]) -> None:
+# @jit(nopython=False, forceobj=True)
+def _pack_strings_old(data: List[str], indices: np.ndarray, strings: List[str], offsets: List[int]) -> None:
     acc_len = 0
     str_map: Dict[str, int] = dict()
 
diff --git a/tests/_run_all_tests.py b/tests/_run_all_tests.py
index fb31d71..9a486c9 100644
--- a/tests/_run_all_tests.py
+++ b/tests/_run_all_tests.py
@@ -1,15 +1,15 @@
 import unittest
 
 testmodules = [
-    "byte_array",
-    "delta",
-    "fixed_point",
-    "integer_packing",
-    "interval_quantization",
-    "run_length",
+    # "byte_array",
+    # "delta",
+    # "fixed_point",
+    # "integer_packing",
+    # "interval_quantization",
+    # "run_length",
     "string_array",
-    "_decoding",
-    "_encoding",
+    # "_decoding",
+    # "_encoding",
 ]
 
 suite = unittest.TestSuite()
diff --git a/tests/benchmark_string_array.py b/tests/benchmark_string_array.py
new file mode 100644
index 0000000..a7b2f81
--- /dev/null
+++ b/tests/benchmark_string_array.py
@@ -0,0 +1,95 @@
+from argparse import ArgumentError
+from ciftools.binary.encoder import STRING_ARRAY
+import pytest
+import msgpack
+import numpy as np
+import random, string
+from timeit import default_timer as timer
+
+MAX_LENGTH = 30
+
+def _random_string(length):
+    return ''.join(random.choice(string.ascii_letters) for i in range(length))
+
+
+# Source:
+def _generate_random_strings_list(size: int, n_of_strings: int):
+    l = [_random_string(random.randint(1, MAX_LENGTH)) for i in range(n_of_strings)]
+    repeated_l = l * size
+    return repeated_l
+
+INPUTS_FOR_ENCODING = {}
+
+for n_of_strings in [20, 50, 100]:
+    for size in [10**4, 10**5, 10**6]:
+        INPUTS_FOR_ENCODING[f'{n_of_strings}_{size}'] = _generate_random_strings_list(size=size, n_of_strings=n_of_strings)
+
+def encoding(encoding_input, optimization):
+    encoder = STRING_ARRAY
+    if not optimization:
+        encoded = encoder.encode_not_optimized(encoding_input)
+    else:
+        encoded = encoder.encode(encoding_input)
+
+print('not_optimized, not_optimized_2nd, not_optimized_3rd, optimized, optimized_2nd, optimized_3rd')
+for case_name, encoding_input in INPUTS_FOR_ENCODING.items():
+
+    start_not_optimized = timer()
+    encoding(encoding_input=encoding_input, optimization=False)
+    stop_not_optimized = timer()
+
+    start_not_optimized_2nd = timer()
+    encoding(encoding_input=encoding_input, optimization=False)
+    stop_not_optimized_2nd = timer()
+
+    start_not_optimized_3rd = timer()
+    encoding(encoding_input=encoding_input, optimization=False)
+    stop_not_optimized_3rd = timer()
+
+    start_optimized = timer()
+    encoding(encoding_input=encoding_input, optimization=True)
+    stop_optimized = timer()
+
+    start_optimized_2nd = timer()
+    encoding(encoding_input=encoding_input, optimization=True)
+    stop_optimized_2nd = timer()
+
+    start_optimized_3rd = timer()
+    encoding(encoding_input=encoding_input, optimization=True)
+    stop_optimized_3rd = timer()
+
+    not_optimized = stop_not_optimized - start_not_optimized
+    not_optimized_2nd = stop_not_optimized_2nd - start_not_optimized_2nd
+    not_optimized_3rd = stop_not_optimized_3rd - start_not_optimized_3rd
+
+    optimized = stop_optimized - start_optimized
+    optimized_2nd = stop_optimized_2nd - start_optimized_2nd
+    optimized_3rd = stop_optimized_3rd - start_optimized_3rd
+
+    print(case_name)
+    print(not_optimized, not_optimized_2nd, not_optimized_3rd, optimized, optimized_2nd, optimized_3rd)
+
+# print(not_optimized, not_optimized_2nd, not_optimized_3rd, optimized, optimized_2nd, optimized_3rd)
+# 0.678212083876133 0.07555452641099691 0.0717478571459651 0.8585679298266768 0.12347683496773243 0.1249676188454032
+# quite close to pytest-benchmark
+
+# NO JIT
+# not_optimized, not_optimized_2nd, not_optimized_3rd, optimized, optimized_2nd, optimized_3rd
+# 20_10000
+# 2.671183445956558 0.037990274955518544 0.038723392062820494 0.02901634795125574 0.027696584002114832 0.027359343017451465
+# 20_100000
+# 0.35828070098068565 0.35529211803805083 0.3570930790156126 0.2672198748914525 0.26997689506970346 0.26407589903101325
+# 20_1000000
+# 3.754518774920143 3.6613519359380007 3.590928812045604 2.7633999809622765 2.742795309983194 2.7856404760386795
+# 50_10000
+# 0.08946576493326575 0.08893638011068106 0.08875931904185563 0.05730705999303609 0.05827989405952394 0.05729172995779663
+# 50_100000
+# 0.886467493022792 0.8783201409969479 0.884077426046133 0.7304035819834098 0.7092675910098478 0.7059366019675508
+# 50_1000000
+# 9.578126006992534 9.592585199978203 9.336758888093755 7.011043209931813 6.888902453938499 6.8985374378971756
+# 100_10000
+# 0.17676137096714228 0.1785169130889699 0.17561967996880412 0.12274663499556482 0.12944400194101036 0.12680783797986805
+# 100_100000
+# 1.7786761060124263 1.7187940969597548 1.7390421900199726 1.4047608590917662 1.4061879760120064 1.3963392629520968
+# 100_1000000
+# 17.475244562956505 17.666481979074888 17.797239091945812 13.829927618033253 13.793060117051937 13.868345592985861
\ No newline at end of file