Two conditional clause count in pandas

I have a df which looks like this:

api_spec_id  type_of_change   label
213              Breaking      NaN
213              Breaking      major
213              Non-Breaking  patch
345              Non-Breaking  NaN
345              Non-Breaking  patch
345              Non-Breaking  patch
678              Breaking      NaN
678              Breaking      minor
678              Breaking      major
123              Breaking      NaN
123              Breaking      NaN

I want to calculate the unique number of api_spec_id where:

all type_of_change are breaking, where expected output would be 2 (ids: 678,123)

all type_of_change are non-breaking, where expected output would be 1 (id:345)

at least one type_of_change is breaking, expected output: 3 (ids:213,678,123)

at least one type_of_change is non-breaking, expected output: 2 (ids:213,345)

I am not sure how I can achieve this, any suggestions or ideas would be greatly appreciated.

A possible solution :

from functools import partial

grp = df.groupby("api_spec_id")

def detect(g, how, change):
    if how == "all":
        return g["type_of_change"].eq(change).all()
    elif how == "any":
        return g["type_of_change"].eq(change).any()

def get_id(df):
    return df["api_spec_id"].unique().tolist()
    
v1 = grp.filter(partial(detect, how="all", change="Breaking")).pipe(get_id)
v2 = grp.filter(partial(detect, how="all", change="Non-Breaking")).pipe(get_id)
v3 = grp.filter(partial(detect, how="any", change="Breaking")).pipe(get_id)
v4 = grp.filter(partial(detect, how="any", change="Non-Breaking")).pipe(get_id)

Output :

print(v1) # [678, 123]
print(v2) # [345]
print(v3) # [213, 678, 123]
print(v4) # [213, 345]

We can do that by grouping by the api_spec_id and label, then applying a lambda function that skips the first element for each group, then we create a new column indicating where label == 'major' and type_of_change == 'Breaking'.

To get the counts of each api_spec_id we can use groupby again, then we can iterate over the unique label values with the same process. Here is the code below, dont forget to replace 'Breaking' with 'Non-Breaking' and other label types.

edit: based on your comments, now we will group by api_spec_id and then check the unique values of type_of_change within each group

import pandas as pd

df = pd.DataFrame({
    "api_spec_id": [213, 213, 213, 345, 345, 345, 678, 678, 678, 123, 123],
    "type_of_change": ["Breaking", "Breaking", "Non-Breaking", "Non-Breaking", 
                       "Non-Breaking", "Non-Breaking", "Breaking", "Breaking", 
                       "Breaking", "Breaking", "Breaking"],
    "label": [None, "major", "patch", None, "patch", "patch", None, "minor", 
              "major", None, None]
})

unique_changes = df.groupby('api_spec_id')['type_of_change'].unique()

all_breaking = unique_changes.apply(lambda x: len(x)==1 and "Breaking" in x).sum()
all_non_breaking = unique_changes.apply(lambda x: len(x)==1 and "Non-Breaking" in x).sum()
at_least_one_breaking = unique_changes.apply(lambda x: "Breaking" in x).sum()
at_least_one_non_breaking = unique_changes.apply(lambda x: "Non-Breaking" in x).sum()

print('All breaking: ', all_breaking)
print('All non-breaking: ', all_non_breaking)
print('At least one breaking: ', at_least_one_breaking)
print('At least one non-breaking: ', at_least_one_non_breaking)

Here is an option:

l = ['None','All Breaking, Has Breaking','All Non Breaking, Has Non Breaking','Has Breaking, Has Non Breaking']
d = df.set_index('api_spec_id')['type_of_change'].str.get_dummies().groupby(level=0).sum().gt(0)
d = (d.mul([1,2])
 .sum(axis=1)
 .map(dict(enumerate(l)))
 .str.split(', ')
 .explode()
 .reset_index(name = 'Type')
 .groupby('Type')['api_spec_id'].agg(list)
 .to_dict())

Output:

{'All Breaking': [123, 678],
 'All Non Breaking': [345],
 'Has Breaking': [123, 213, 678],
 'Has Non Breaking': [213, 345]}