数据类型“国家/地区”无法理解

Question

I am getting the following error for my code: data type "country" not understood. I am relatively new to python and am basically trying to learn how to work with .csv files. I'm using python 3.4 and editor Canopy. I was trying to format the data types of the csv into strings and floats, but as soon as i try to assign string type to the first data column (the col is headed by the word - country) i get the error. I am trying to assign country to "a200" type which is believe can be a string. What am i doing wrong here? Please be clear as i am new.

The code is this:

import csv
import numpy 

def open_with_csv(filename):

    data = []

    with open(filename) as csvin:
        file_reader = csv.reader(csvin, delimiter = ',')
        for line in file_reader:
            data.append(line)

    return data

data_from_csv = open_with_csv('C:\Users\user\Desktop\MDR-TB_burden_estimates_2015-05-07.csv')

print (data_from_csv)

FIELDNAMES = ['country', 'iso2', 'iso3', 'iso_numeric', 'year', 'source_mdr_new', 'source_drs_coverage_new', 'source_drs_year_new', 'e_new_mdr_pcnt', 'e_new_mdr_pcnt_lo', 'e_new_mdr_pcnt_hi', 'e_new_mdr_num', 'e_new_mdr_num_lo', 'e_new_mdr_num_hi', 'source_mdr_ret', 'source_drs_coverage_ret', 'source_drs_year_ret', 'e_ret_mdr_pcnt', 'e_ret_mdr_pcnt_lo', 'e_ret_mdr_pcnt_hi', 'e_ret_mdr_num', 'e_ret_mdr_num_lo', 'e_ret_mdr_num_hi', 'e_mdr_num', 'e_mdr_num_lo', 'e_mdr_num_hi']


print (FIELDNAMES)

DATATYPES = [('country','a200'), ('iso2'), ('iso3'), ('iso_numeric'), ('year'), ('source_mdr_new'), ('source_drs_coverage_new'), ('source_drs_year_new'), ('e_new_mdr_pcnt'), ('e_new_mdr_pcnt_lo'), ('e_new_mdr_pcnt_hi'), ('e_new_mdr_num'), ('e_new_mdr_num_lo'), ('e_new_mdr_num_hi'), ('source_mdr_ret'), ('source_drs_coverage_ret'), ('source_drs_year_ret'), ('e_ret_mdr_pcnt'), ('e_ret_mdr_pcnt_lo'), ('e_ret_mdr_pcnt_hi'), ('e_ret_mdr_num'), ('e_ret_mdr_num_lo'), ('e_ret_mdr_num_hi'), ('e_mdr_num'), ('e_mdr_num_lo'), ('e_mdr_num_hi')]

def load_data(filename, d=','):
    my_csv = numpy.genfromtxt(filename, delimiter=d, skip_header=1, invalid_raise=False, names= FIELDNAMES, dtype = DATATYPES)
    return my_csv

my_csv = load_data('C:\Users\user\Desktop\MDR-TB_burden_estimates_2015-05-07.csv')

Answer 1

Looks like the arguments you are passing to numpy.getfromtxt are incorrectly formatted.

If you want to pass a value to both names and dtype arguments then you need to specify dtype as a coma separated string: "a200, i4, etc..."

Alternatively you can pass a list of tuple ("name", "type") pairs and not specify names argument.

You can look here for examples: http://docs.scipy.org/doc/numpy/reference/generated/numpy.genfromtxt.html

Answer 2

I believe this recreates your problem:

In [156]: txt=b"""USA, 123, ux345, 1.24
Canada, 434, xz3444, 3.34
France, 443, 2x453, 4.34
"""    
In [157]: FIELDNAMES=['country','id','code','value']
In [158]: DATATYPES =[('country','a100'),('id'),('code'),('value')]
In [159]: np.genfromtxt(txt.splitlines(), delimiter=',',dtype=DATATYPES, names=FIELDNAMES)

---

...
--> 847         ndtype = np.dtype(dict(formats=ndtype, names=names))
    848     else:
    849         nbtypes = len(ndtype)

TypeError: data type "country" not understood

So that's where your subject line is coming from. For some reason in parsing DATATYPES it thinks 'country' denotes a dtype (on the par with 'S100', 'int', etc). But you mean it to be a field name.

Let's correct DATATYPES , and supply a type for each of the fields, not just the first:

In [165]: DATATYPES =[('country','a100'),('id',int),('code','a5'),('value',float)]

In [166]: np.genfromtxt(txt.splitlines(), delimiter=',',dtype=DATATYPES, names=FIELDNAMES)
Out[166]: 
array([(b'USA', 123, b' ux34', 1.24), (b'Canada', 434, b' xz34', 3.34),
       (b'France', 443, b' 2x45', 4.34)], 
      dtype=[('country', 'S100'), ('id', '<i4'), ('code', 'S5'), ('value', '<f8')])

As Serguei writes, there are several ways of specifying the names and dtypes. So yes, reread the genfromtxt docs if this isn't clear. There are also a lot of genfromtxt questions and examples on SO.