Loop through a dictionary, reference key

Question

I'm trying to loop through a dictionary, but use the looped keys (or values if I so choose) in a calculation.

EDIT 11/24: Adding additional context

So the example would be:

STEP 1) I have a CSV file with the following columns... Column 1, Column 2, Column 3

file = pd.read_csv("/Path/test.csv")

time   Change    Column 1   Column 2   Column 3   Column 4
01/01    .5          3          5          7         1
01/01    .3          5          1          4         2
01/01    .5          1          3          8         2
01/02    .3          5          1          4         1
01/02    .5          1          3          8         3

STEP 2) I have a dictionary that groups these columns and provides information regarding the calculations required for the columns in Key Value pairs. I am additionally assigning a code to each column for use later (NOTE: There are 10 groups of columns, containing anywhere from 4-20 columns per group, below is simply an example):

dict = {
'group': {
    'a':{
       'ranges':{
             'a':{'<':0,'>':-1},
             'b':{'<':0,'>':-2}} 
       'indicators':{
             'a':'Column 1',
             'b':'Column 2'}
    'b':{
       'ranges':{
             'a':{'<':0,'>':-1},
             'b':{'<':0,'>':-2}} 
       'indicators':{
             'a':'Column 3',
             'b':'Column 4'}

STEP 3 [This is where I'm having trouble]) The 'Groups' are a collection of columns with 'similar' data as well as corresponding ranges I want to test against. I am trying to compare different combinations of groups with other groups. So compare ALL the different RANGES of Column 1 to all the different RANGES of Column 3, then 4... then compare ALL the different RANGES of Column 2 to all the different RANGES of Column 3, then 4. My thought was to assign a CODE based on the corresponding letters for GROUP, RANGES, and INDICATORS... so for example "aaa" would be GROUP a, RANGES {'<':0,'>':-1}, in INDICATOR Column 1

NOTE: I only need to compare 3 columns at a time, e.g. the loop can stop at 3 combinations, however I'd like to also know if 2 combinations are better than 3.

STEP 4) I want to group the different combinations to see which combinations of COLUMNS and RANGES work the best, by calculating the MEAN, MAX, MIN, and then COUNT of each combination.

I am attempting to create a huge loop for this, but in an effort to learn on my own I'm only asking the community pieces of it. But I'm seeing that this piecemeal approach isn't helpful to understand the context of my issue. So hopefully this additional explanation provides a bit more clarity

My eventual desired output:

Code   Average   Max   Min   Count
aaa    0.25      0.5   0.3   3
aab    0.25      0.5   0.3   3
aac    0.25      0.5   0.3   3
aba    0.25      0.5   0.3   3
abb    0.25      0.5   0.3   3
abc    0.25      0.5   0.3   3

Answer 1

Are you just having trouble with iterating over nested dictionaries and accessing their keys and values? Something like this?

for group_name, columns in dict['group'].items():
    for col, comparisons in columns.items():
        for operator, number in comparisons.items():
            print(operator, number)

UPDATE:

If you want to create the code to be executed using the retrieved operators and numbers, one approach that I can think of is to use eval.

First, format the string, inserting the column name, operator and number:

f"df['{col}'] {operator} {number}"
# result: "df['Column 6'] > 6"

You can then evaluate this string using the built-in Python eval function:

eval(f"df['{col}'] {operator} {number}")

Note that this will not print anything; if you wanted to see the output, you'd have to add print:

eval(f"print(df['{col}'] {operator} {number})")
# This should print something like:
# index1 False
# index2 True
# index3 False
# Name: Column 6, dtype: bool

The entire code could look like this:

for group_name, columns in dict["group"].items():
    for col, comparisons in columns.items():
        for operator, number in comparisons.items():
            column_fulfils_condition = eval(f"df['{col}'] {operator} {number}")  
            if column_fulfils_condition["index1"]:
                print(f"Yay! index1 for {col} is {operator} than {number})  # or do whatever else you need

eval can only handle very simple code (single expressions), so it may not be enough for your use case. There are other approaches, which have been discussed elsewhere.

Answer 2

If I understand you correctly, you would like to turn the string representation of an operator into an operator you can apply to the values in the dictionary (is that correct?) In that case you could, for example, define a dictionary mapping the symbol to the corresponding operator from the module operator,

import operator
ops = {'<': operator.lt,
       '<=': operator.le,
       '=': operator.eq,
       '>': operator.gt,
       '>=': operator.ge,
       '!=': operator.ne
      }

and then receive this operator from its symbol s via ops[s], e.g.

s= '<'
ops[s](1, 2)  # evaluates to True