How to implement key of key in python?

[Peter Otten]

eckhleung at gmail.com wrote:

> On Saturday, May 10, 2014 10:30:06 AM UTC+8, MRAB wrote:
>> On 2014-05-10 02:22, I wrote:
>> 
>> > I'm migrating from Perl to Python and unable to identify the equivalent
>> > of key of key concept. The following codes run well,
>> 
>> > import csv
>> 
>> > attr = {}
>> 
>> > with open('test.txt','rb') as tsvin:
>> 
>> >      tsvin = csv.reader(tsvin, delimiter='\t')
>> 
>> >      for row in tsvin:
>> 
>> >          ID = row[1]
>> 
>> > until:
>> 
>> >          attr[ID]['adm3'] = row[2]
>> 
>> > I then try:
>> 
>> >          attr[ID].adm3 = row[2]
>> 
>> > still doesn't work. Some posts suggest using module dict but some do
>> > not. I'm a bit confused now. Any suggestions?
>> 
>> Python doesn't have Perl's autovivication feature. If you want the
>> 
>> value to be a dict then you need to create that dict first:
>> 
>> attr[ID] = {}
>> 
>> attr[ID]['adm3'] = row[2]
>>
>> You could also have a look at the 'defaultdict' class in the
>> 
>> 'collections' module.
> 
> I identify the information below:
> s = [('yellow', 1), ('blue', 2), ('yellow', 3), ('blue', 4), ('red', 1)]
> d = defaultdict(list)
> for k, v in s:
>   d[k].append(v)
> 
> While it is fine for a small dataset, I need a more generic way to do so.
> Indeed the "test.txt" in my example contains more columns of attributes
> like:
> 
> ID address age gender phone-number race education ...
> ABC123 Ohio, USA 18 F 800-123-456 european university
> ACC499 London 33 M 800-111-400 african university
> ...
> 
> so later I can retrieve the information in python by:
> 
> attr['ABC123'].address (containing 'Ohio, USA')
> attr['ABC123'].race (containing 'european')
> attr['ACC499'].age (containing '33')

Using a csv.DictReader comes close with minimal effort:

# write demo data to make the example self-contained
with open("tmp.csv", "w") as f:
    f.write("""\
ID,address,age,gender,phone-number,race,education
ABC123,"Ohio, USA",18,F,800-123-456,european,university
ACC499,London,33,M,800-111-400,african,university
""")

import csv
import pprint

with open("tmp.csv") as f:
    attr = {row["ID"]: row for row in csv.DictReader(f)}
        
pprint.pprint(attr)

print(attr["ACC499"]["age"])

The "dict comprehension"

    attr = {row["ID"]: row for row in csv.DictReader(f)}

is a shortcut for

attr = {}
for row in csv.DictReader(f):
    attr[row["ID"]] = row

If you insist on attribute access (row.age instead of row["age"]) you can 
use a namedtuple. This is a bit more involved:

import csv
import pprint
from collections import namedtuple

with open("tmp.csv") as f:
    rows = csv.reader(f)
    header = next(rows)

    # make sure column names are valid Python identifiers
    header = [column.replace("-", "_") for column in header]

    RowType = namedtuple("RowType", header)
    key_index = header.index("ID")
    attr = {row[key_index]: RowType(*row) for row in rows}

pprint.pprint(attr)

print(attr["ABC123"].race)

> The following links mention something similar,

Too many, so I checked none of them ;)