[bug] fix numpy 1d array length encoding bug, fix #4

Author: fangq

bjdata/encoder.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2022 Qianqian Fang <q.fang at neu.edu>. All rights reserved.
+# Copyright (c) 2020-2023 Qianqian Fang <q.fang at neu.edu>. All rights reserved.
 # Copyright (c) 2016-2019 Iotic Labs Ltd. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -288,13 +288,10 @@ def __encode_numpy(fp_write, item, islittle, default):
         item = np.array( item, order = 'C')  # currently, BJData ND-array syntax only support row-major
 
     fp_write(ARRAY_START + CONTAINER_TYPE + __map_dtype(item.dtype.str) + CONTAINER_COUNT)
-    if item.ndim == 1:
-        __encode_int(fp_write, len(item), islittle)
-    else:
-        fp_write(ARRAY_START)
-        for value in item.shape:
-            __encode_int(fp_write, value, islittle)
-        fp_write(ARRAY_END)
+    fp_write(ARRAY_START)
+    for value in item.shape:
+        __encode_int(fp_write, value, islittle)
+    fp_write(ARRAY_END)
 
     fp_write(item.data)
 

src/encoder.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2022 Qianqian Fang <q.fang at neu.edu>. All rights reserved.
+ * Copyright (c) 2020-2023 Qianqian Fang <q.fang at neu.edu>. All rights reserved.
  * Copyright (c) 2016-2019 Iotic Labs Ltd. All rights reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
@@ -317,16 +317,14 @@ static int _encode_NDarray(PyObject *obj, _bjdata_encoder_buffer_t *buffer) {
         WRITE_CHAR_OR_BAIL((char)marker);
     }
     WRITE_CHAR_OR_BAIL(CONTAINER_COUNT);
-    if(ndim == 1) {
-        BAIL_ON_NONZERO(_encode_longlong(bytes, buffer));
-    } else {
-        WRITE_CHAR_OR_BAIL(ARRAY_START);
-        for(int i=0 ; i<ndim; i++)
-            _encode_longlong(dims[i], buffer);
-        if(type == NPY_UNICODE)
-            _encode_longlong(4, buffer);
-        WRITE_CHAR_OR_BAIL(ARRAY_END);
-    }
+
+    WRITE_CHAR_OR_BAIL(ARRAY_START);
+    for(int i=0 ; i<ndim; i++)
+        _encode_longlong(dims[i], buffer);
+    if(type == NPY_UNICODE)
+        _encode_longlong(4, buffer);
+    WRITE_CHAR_OR_BAIL(ARRAY_END);
+
     WRITE_OR_BAIL(PyArray_BYTES(arr), bytes*total);
     Py_DECREF(arr);
     // no ARRAY_END since length was specified

test/test.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2022 Qianqian Fang <q.fang at neu.edu>. All rights reserved.
+# Copyright (c) 2020-2023 Qianqian Fang <q.fang at neu.edu>. All rights reserved.
 # Copyright (c) 2016-2019 Iotic Labs Ltd. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -314,6 +314,13 @@ def test_nd_array(self):
         self.assertEqual((self.bjdloadb(self.bjddumpb(np.float16(2.2)))  == 16486), True)
         self.assertEqual((self.bjdloadb(self.bjddumpb(np.float32(2.2)))  == np.float32(2.2)), True)
 
+        self.assertEqual((self.bjdloadb(self.bjddumpb(np.array([1.3,-0.5,0.7,1000,11], dtype=np.float32))) == np.array([1.3,-0.5,0.7,1000,11], dtype=np.float32)).all(), True)
+        self.assertEqual((self.bjdloadb(self.bjddumpb(np.array([1,2,3,4], dtype=np.uint16))) == np.array([1,2,3,4], dtype=np.uint16)).all(), True)
+        self.assertEqual((self.bjdloadb(self.bjddumpb(np.array([1,2,3,4], dtype=np.uint8))) == np.array([1,2,3,4], dtype=np.uint8)).all(), True)
+        self.assertEqual((self.bjdloadb(self.bjddumpb(np.array([], dtype=np.int8))) == np.array([], dtype=np.int8)).all(), True)
+        self.assertEqual((self.bjdloadb(self.bjddumpb(np.array([-1,-2,-3,-4], dtype=np.float32))) == np.array([-1,-2,-3,-4], dtype=np.float32)).all(), True)
+        self.assertEqual((self.bjdloadb(self.bjddumpb(np.array([[-1,-2,5],[-3,-4,-6]], dtype=np.float64))) == np.array([[-1,-2,5],[-3,-4,-6]], dtype=np.float64)).all(), True)
+
         raw_start = (ARRAY_START + CONTAINER_TYPE + TYPE_INT8 + CONTAINER_COUNT + ARRAY_START + \
                     TYPE_UINT8 + b'\x03' + TYPE_UINT16 + b'\x02' + b'\x00' + ARRAY_END + \
                     b'\x01'+ b'\x02'+ b'\x03'+ b'\x04'+ b'\x05'+ b'\x06')