The following two approaches can be utilized in order to create a custom message that contains a 2D array, (map_data in your case):

1. Using nested .msg: In this approach, you define two .msg files as shown below:

   1. Create Row.msg file with the following content:

      int8[] data
      

      • Here, you create a row of your map.
      • Please note that, I am assuming that map contains data of type int8. Please change it as per your need.

   2. Create Map.msg file with the following content:

      Row[] row
      

      • Here, you append the above created row to your msg.
      • Please check the following link for more info. #q67273
2. Using flattened array: In this approach, you reshape your 2D array to make it a 1D array. This is the simplest approach and used by ROS to publish images (2D data). You are free to check the message definition of sensor_msgs/Image and read more about it.

Example

Consider you have the following 2D array:

In [1]: map_data
Out[1]: 
array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8,  9, 10, 11]], dtype=int8)

1. Approach #1: Create every row and append it to map as shown below:

   # initialize map
   map = Map()
   
   # initialize row with 1D data
   row1 = Row([ 0,  1,  2,  3])
   
   # append row to map
   map.row.append(row1)
   
   row2 = Row([ 4,  5,  6,  7])
   map.row.append(row2)
   row3 = Row([ 8,  9, 10, 11])
   map.row.append(row3)
   

   1. NumPy Simulation:

      In [2]: row1 = np.array([ 0,  1,  2,  3])
      In [3]: row2 = np.array([ 4,  5,  6,  7])
      In [4]: row3 = np.array([ 8,  9, 10, 11])
      
      In [5]: map_data_received = np.vstack((row1, row2, row3))
      In [6]: map_data_received
      Out[6]: 
      array([[ 0,  1,  2,  3],
             [ 4,  5,  6,  7],
             [ 8,  9, 10, 11]])
      

2. Approach #2: Just flatten your data and publish it. Upon subscribing, reshape it back to 2D as shown in the NumPy simulation below:

   In [6]: map_data.flatten()
   Out[6]: array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11])
   
   In [7]: map_data_received = map_data.flatten().reshape((3,-1))
   In [8]: map_data_received
   Out[8]: 
   array([[ 0,  1,  2,  3],
          [ 4,  5,  6,  7],
          [ 8,  9, 10, 11]])
   

On a side note, it seems that your variable map_data_values is already a 1D array. So, approach #2 can be applied directly.