Hey there, let's break down why you're hitting that pandas._libs.hashtable.Int64HashTable.get_item error and fix it up.

First, looking at your code and output, the root issue is chained indexing—that's when you do something like df[df.result == 'Won'][df['my_classification'] == c]['prob'][0].

Why this causes problems #

Pandas can return either a view or a copy of your original DataFrame when you use chained indexing, and that ambiguity messes up subsequent operations. Sometimes you end up trying to index into a copy that doesn't have the structure you expect, leading to that hashtable error you're seeing.

Luckily, your data has exactly one row per (result, my_classification) pair, so we can fix this cleanly with safer indexing methods.

Fixed Code Option 1: Use .loc for safe combined indexing #

.loc is Pandas' recommended way to index rows and columns together, avoiding view/copy confusion:

print(df)
categories = df['my_classification'].unique()
for c in categories:
    print(c)
    # Combine conditions in .loc to get the exact value
    win = df.loc[(df['result'] == 'Won') & (df['my_classification'] == c), 'prob'].iloc[0]
    print(type(win))
    lost = df.loc[(df['result'] == 'Lost') & (df['my_classification'] == c), 'prob'].iloc[0]
    print(type(lost))

Fixed Code Option 2: Use groupby for cleaner (and faster) results #

If you're working with larger datasets later, grouping first will avoid repeated filtering in the loop:

print(df)
# Group by classification and result, grab the first prob value per group
grouped_probs = df.groupby(['my_classification', 'result'])['prob'].first()

categories = df['my_classification'].unique()
for c in categories:
    print(c)
    win = grouped_probs.loc[c, 'Won']
    print(type(win))
    lost = grouped_probs.loc[c, 'Lost']
    print(type(lost))

Key Notes #

• Parentheses around conditions: When combining boolean filters with &, always wrap each condition in parentheses—otherwise Python's operator priority will mess things up.
• .iloc[0] for scalar values: Since each filter matches exactly one row, .iloc[0] pulls out the float value instead of returning a Series.
• Avoid chained indexing: Stick to .loc (or .iloc for positional indexing) whenever you can to prevent these hard-to-debug errors.

Expected Output #

Running either fixed code with your sample data will give you this:

result my_classification      prob
0    Won        ENTERPRISE  0.657895
1    Won       COMMERCIAL  0.342105
2   Lost        ENTERPRISE  0.611842
3   Lost       COMMERCIAL  0.388158
ENTERPRISE
<class 'float'>
<class 'float'>
COMMERCIAL
<class 'float'>
<class 'float'>

内容的提问来源于stack exchange，提问作者Edamame