Array slicing with data that contains multiple types

Question

I read in the following object with numpy.genfromtxt:

A = [(4, 'A', 3750.5), 
     (4, 'B', 3252.6),
     (8, 'A', 3350.5), 
     (8, 'B', 3152.6)]

I would like to do numpy fancy indexing on it, but I can't because this is not an numpy array. It's an array of a list.

What would be the best way to get the 3rd column of all rows that have '4' in the first column?

I tried A[A[:,0]==4] but the interpreter complained with "IndexError: invalid index".

Edit:

This is the python program I am using:

import numpy as np

A = np.genfromtxt( "text.txt" , dtype=( int , "|S10", float))

A_array = np.asarray(A, dtype=object)

print A
print A_array

The file text.txt:

This is the output:

[(4, 'A', 3750.5) (4, 'B', 3270.5) (8, 'A', 3480.5) (8, 'B', 3590.5)]
[(4, 'A', 3750.5) (4, 'B', 3270.5) (8, 'A', 3480.5) (8, 'B', 3590.5)]

What am I missing here?

Why did someone insert commas into A? If I print A there are no commas. Is it maybe a list of tuples? — ritter
– ritter, Commented Nov 15, 2013 at 21:26
You used A = ... implying that you're assigning the data. When you "print" data, you don't say A = ..., you just print the contents. When you assign to a variable with A = ... you need the value on the right-hand-side of the = symbol to evaluate to a valid Python object, which is not true without the commas. — ely
– ely, Commented Nov 15, 2013 at 21:29
I will try to insert commas as needed. I explained it the wrong way. The first matrix was the output of a print command. The question is updated. Sorry — ritter
– ritter, Commented Nov 15, 2013 at 21:36
Ok. Do you agree that A_array and A would print exactly the same? — ritter
– ritter, Commented Nov 15, 2013 at 21:39

ely · Accepted Answer · 2013-11-15 21:16:56Z

4

In [24]: A_array = numpy.asarray(A, dtype=object)

In [25]: A_array[A_array[:,0] == 4]
Out[25]:
array([[4, A, 3750.5],
       [4, B, 3252.6]], dtype=object)

If the columns of data have semantic meaning that you'd like to keep track of, consider loading the list of tuples directly into a Pandas DataFrame and giving them column labels. The logical indexing would work similarly:

In [27]: A_df = pandas.DataFrame(A, columns=['Col1', 'Col2', 'Col3'])

In [28]: A_df
Out[28]:
   Col1 Col2    Col3
0     4    A  3750.5
1     4    B  3252.6
2     8    A  3350.5
3     8    B  3152.6

In [29]: A_df.Col1 == 4
Out[29]:
0     True
1     True
2    False
3    False
Name: Col1

In [30]: A_df[A_df.Col1 == 4]
Out[30]:
   Col1 Col2    Col3
0     4    A  3750.5
1     4    B  3252.6

edited Nov 15, 2013 at 21:16

answered Nov 15, 2013 at 21:10

ely

77.8k36 gold badges158 silver badges234 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

ritter Over a year ago

Hmm.. I tried your first solution. When I print A and A_array they appear exactly the same, and the indexing command on A_array fails as it does on A.

ely Over a year ago

Your comment is ambiguous unless I can see the specific code you're executing. It works for me, and is indeed the standard solution for this in numpy.

ely Over a year ago

Also, if A and A_array appear the same when you use print to print them out, you must have another error. The first, being just a Python list will print to the console very differently than the second, which is a numpy.ndarray. For instance, A_array will always some dtype field printed along with its actual data contents, whereas that doesn't exist for a list.

DaoWen · Accepted Answer · 2013-11-15 21:12:08Z

2

First, you need commas between the list elements in A, otherwise you'll get a syntax error:

A = [(4, 'A', 3750.5),
     (4, 'B', 3252.6),
     (8, 'A', 3350.5), 
     (8, 'B', 3152.6)]

Next, you can use a list comprehension to get what you want pretty succinctly:

[ row[2] for row in A if row[0] == 4 ]

Result:

[3750.5, 3252.6]

answered Nov 15, 2013 at 21:12

DaoWen

33.1k6 gold badges77 silver badges102 bronze badges

Collectives™ on Stack Overflow

Array slicing with data that contains multiple types

2 Answers 2

3 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related