Python。 将字符串转换为包含 numpy arrays 的字典

Question

I have a string which has this structure:

{0: [array([5.1, 3.5, 1.4, 0.2]), array([4.9, 3. , 1.4, 0.2]), 1: [array([7. , 3.2, 4.7, 1.4]), array([6.4, 3.2, 4.5, 1.5]), 2: [array([6.3, 3.3, 6. , 2.5]), array([7.1, 3. , 5.9, 2.1])]}

It is in the form of a python dictionary containing numpy arrays. How can I turn this string into a python dictionary containing numpy arrays?

Thanks for the replies already. I can't simply just change the string because the string is an output of a python file that I want to pipe to another one. I could go and change the format of this output, but I was rather hoping to avoid that.

Answer 1

In an interactive ipython session is created an alias: In [7]: array = np.array

and edited a copy-n-paste of your sample. I had to add a couple of closing ] :

In [8]: data = {0: [array([5.1, 3.5, 1.4, 0.2]), array([4.9, 3. , 1.4, 0.2])], 1
   ...: : [array([7. , 3.2, 4.7, 1.4]), array([6.4, 3.2, 4.5, 1.5])], 2: [array(
   ...: [6.3, 3.3, 6. , 2.5]), array([7.1, 3. , 5.9, 2.1])]}

Now a simple execute works:

In [9]: data
Out[9]: 
{0: [array([5.1, 3.5, 1.4, 0.2]), array([4.9, 3. , 1.4, 0.2])],
 1: [array([7. , 3.2, 4.7, 1.4]), array([6.4, 3.2, 4.5, 1.5])],
 2: [array([6.3, 3.3, 6. , 2.5]), array([7.1, 3. , 5.9, 2.1])]}

or with a string (corrected):

In [10]: astr="{0: [array([5.1, 3.5, 1.4, 0.2]), array([4.9, 3. , 1.4, 0.2])], 1
    ...: : [array([7. , 3.2, 4.7, 1.4]), array([6.4, 3.2, 4.5, 1.5])], 2: [array
    ...: ([6.3, 3.3, 6. , 2.5]), array([7.1, 3. , 5.9, 2.1])]}"
In [11]: astr
Out[11]: '{0: [array([5.1, 3.5, 1.4, 0.2]), array([4.9, 3. , 1.4, 0.2])], 1: [array([7. , 3.2, 4.7, 1.4]), array([6.4, 3.2, 4.5, 1.5])], 2: [array([6.3, 3.3, 6. , 2.5]), array([7.1, 3. , 5.9, 2.1])]}'

I can do an exec or eval :

In [15]: eval(astr)
Out[15]: 
{0: [array([5.1, 3.5, 1.4, 0.2]), array([4.9, 3. , 1.4, 0.2])],
 1: [array([7. , 3.2, 4.7, 1.4]), array([6.4, 3.2, 4.5, 1.5])],
 2: [array([6.3, 3.3, 6. , 2.5]), array([7.1, 3. , 5.9, 2.1])]}

Use of eval is often discouraged because it can be hacked. But the safer ast_eval only works with dict, list and tuples, not np.array . One way or other you have to edit the string so it is a valid Python/numpy expression.

Answer 2

I changed how the string was being being produced so that it was in the form of a dictionary containing python lists instead of numpy arrays. With this I could just use ast.literal_eval() to deseriealize it.