framecheck 0.3.0

Lightweight, readable, and extensible validation for pandas DataFrames.

Lightweight, flexible, and intuitive validation for pandas DataFrames.
Define expectations for your data, validate them cleanly, and surface friendly errors or warnings — no configuration files, ever.

---

📦 Installation

pip install framecheck

---

Main Features

• Designed for pandas users
• Simple, fluent API
• Supports error or warning-level assertions
• Validates both column-level and DataFrame-level rules
• No config files, decorators, or boilerplate

---

Table of Contents

• Getting Started
• Output
• Comparison with Other Approaches
  • great_expectations
  • Manual Validation
• FrameCheck Methods
  • .column(...)
  • .columns(...)
  • .columns_are(...)
  • .empty()
  • .not_empty()
  • .only_defined_columns()
  • .row_count(...)
  • .unique(...)
• License
• Contact

---

🔥 Example: Catch data issues before they cause bugs

Example dataframe:

import pandas as pd
from framecheck import FrameCheck

df = pd.DataFrame({
    'a': [0, 1, 0, 1, 2],
    'b': [1, 1, 0, 0, 3],
    'timestamp': ['2022-01-01', '2022-01-02', '2019-12-31', '2021-01-01', '2023-05-01'],
    'email': ['[e-mail]', 'bad', '[e-mail]', 'not-an-email', '[e-mail]'],
    'extra': ['x'] * 5
})

With FrameCheck:

validator = (
    FrameCheck()
    .columns(['a', 'b'], type='int', in_set=[0, 1])
    .column('timestamp', type='datetime', after='2020-01-01')
    .column('email', type='string', regex=r'.+@.+\..+', warn_only=True)
    .only_defined_columns()
    .row_count(min=5, max=100)
    .not_empty()
    .raise_on_error()
)

result = validator.validate(df)

• .warn_only=True allows specific checks to issue warnings instead of failing validation.
• .raise_on_error() makes the entire validation raise an exception if any non-warning failure occurs.
• Together, they let you enforce hard rules while still being lenient on others.

For example, in the code above:

• Invalid email formats will trigger a warning but not block execution.
• Everything else (bad types, missing columns, out-of-bound values, etc.) will raise an error.

🧾 Output

If the data is invalid, you'll get warning ...

FrameCheckWarning: FrameCheck validation warnings:
- Column 'email' has values not matching regex '.+@.+\..+': ['bad'].
  result = validator.validate(df)

... and because you used .raise_on_error(), it'll raise a clean exception:

ValueError: FrameCheck validation failed:
Column 'a' is missing.
Column 'b' is missing.
Column 'timestamp' is missing.
DataFrame must have at least 5 rows (found 3).
Unexpected columns in DataFrame: ['good_credit', 'home_owner', 'id', 'promo_eligible', 'score']

---

Comparison with Other Approaches

Equivalent code in great_expectations (which is a fantastic package with a much broader focus than FrameCheck).

import great_expectations as ge

ge_df = ge.from_pandas(df)
ge_df.expect_column_values_to_be_in_set('a', [0, 1])
ge_df.expect_column_values_to_be_of_type('a', 'int64')
ge_df.expect_column_values_to_be_in_set('b', [0, 1])
ge_df.expect_column_values_to_be_of_type('b', 'int64')
ge_df['timestamp'] = pd.to_datetime(ge_df['timestamp'])
ge_df.expect_column_values_to_be_of_type('timestamp', 'datetime64[ns]')
ge_df.expect_column_values_to_be_between('timestamp', min_value='2020-01-01')
ge_df.expect_column_values_to_match_regex('email', r'.+@.+\..+', mostly=1.0)
ge_df.expect_table_row_count_to_be_between(min_value=5, max_value=100)
ge_df.expect_table_row_count_to_be_greater_than(0)
expected_columns = {'a', 'b', 'timestamp', 'email'}
unexpected = set(df.columns) - expected_columns
if unexpected:
    raise ValueError(f"Unexpected columns in DataFrame: {unexpected}")

results = ge_df.validate()
if not results['success']:
    raise ValueError(f"Validation failed: {results}")

Equivalent code without a package:

import pandas as pd
import re

errors = []

if df.empty:
    errors.append("DataFrame is empty.")

row_count = len(df)
if row_count < 5:
    errors.append("DataFrame has fewer than 5 rows.")
if row_count > 100:
    errors.append("DataFrame has more than 100 rows.")

for col in ['a', 'b']:
    if col not in df.columns:
        errors.append(f"Missing column: {col}")
    else:
        if not pd.api.types.is_integer_dtype(df[col]):
            errors.append(f"Column '{col}' is not of integer type.")
        if not df[col].isin([0, 1]).all():
            errors.append(f"Column '{col}' contains values outside [0, 1].")

if 'timestamp' not in df.columns:
    errors.append("Missing column: 'timestamp'")
else:
    try:
        ts = pd.to_datetime(df['timestamp'], errors='coerce')
        if ts.isna().any():
            errors.append("Column 'timestamp' contains non-datetime values.")
        elif (ts < pd.Timestamp('2020-01-01')).any():
            errors.append("Column 'timestamp' has values before 2020-01-01.")
    except Exception:
        errors.append("Could not convert 'timestamp' to datetime.")

if 'email' in df.columns:
    invalid_emails = df[~df['email'].astype(str).str.match(r'.+@.+\..+')]
    if not invalid_emails.empty:
        print("Warning: Some emails don't match expected pattern.")

expected_cols = {'a', 'b', 'timestamp', 'email'}
actual_cols = set(df.columns)
unexpected = actual_cols - expected_cols
if unexpected:
    errors.append(f"Unexpected columns in DataFrame: {sorted(unexpected)}")

# Final decision
if errors:
    raise ValueError("Validation failed:\n" + "\n".join(errors))

---

FrameCheck Methods

import pandas as pd
from framecheck import FrameCheck

column(...) – Core Behaviors

✅ Ensures column exists

df = pd.DataFrame({'x': [1, 2, 3]})

schema = FrameCheck().column('x')
result = schema.validate(df)

FrameCheck validation passed.

✅ Type enforcement

df = pd.DataFrame({'x': [1, 2, 'bad']})

schema = FrameCheck().column('x', type='int')
result = schema.validate(df)

FrameCheck validation errors:
- Column 'x' contains values that are not integer-like (e.g., decimals or strings): ['bad'].

---

.column(..., in_set=...) – Allowed values

df = pd.DataFrame({'status': ['new', 'active', 'archived']})

schema = FrameCheck().column('status', in_set=['new', 'active'])
result = schema.validate(df)

FrameCheck validation errors:
- Column 'status' contains values not in allowed set: ['archived'].

---

.column(..., equals=...) – All values must equal one thing

df = pd.DataFrame({'is_active': [True, False, True]})

schema = FrameCheck().column('is_active', type='bool', equals=True)
result = schema.validate(df)

FrameCheck validation errors:
- Column 'is_active' must equal True, but found values: [False].

---

.column(..., not_null=...) – All values non-null if set to True

df = pd.DataFrame({'is_active': [True, False, None]})

schema = FrameCheck().column('is_active', type='bool', not_null=True)
result = schema.validate(df)

FrameCheck validation errors:
- Column 'is_active' contains missing values.
  result = schema.validate(df)

---

.column(..., regex=...) – Pattern matching (for strings)

df = pd.DataFrame({'email': ['[e-mail]', 'bademail']})

schema = FrameCheck().column('email', type='string', regex=r'.+@.+\..+')
result = schema.validate(df)

FrameCheck validation errors:
- Column 'email' has values not matching regex '.+@.+\..+': ['bademail'].

Go to Top

.column(..., min=..., max=..., after=..., before=...) – Range & bound checks

df = pd.DataFrame({
    'age': [25, 17, 101],
    'score': [0.9, 0.5, 1.2],
    'signup_date': ['2021-01-01', '2019-12-31', '2023-05-01'],
    'last_login': ['2020-01-01', '2026-01-01', '2023-06-15']
})

schema = (
    FrameCheck()
    .column('age', type='int', min=18, max=99)
    .column('score', type='float', min=0.0, max=1.0)
    .column('signup_date', type='datetime', after='2020-01-01', before='2025-01-01')
    .column('last_login', type='datetime', min='2020-01-01', max='2025-01-01')
)
result = schema.validate(df)

FrameCheck validation errors:
- Column 'age' has values less than 18.
- Column 'age' has values greater than 99.
- Column 'score' has values greater than 1.0.
- Column 'signup_date' violates 'after' constraint: 2020-01-01.
- Column 'last_login' violates 'max' constraint: 2025-01-01.

Go to Top

columns(...)

Any .column() operation can be applied to multiple columns of the same type.

df = pd.DataFrame({
    'a': [0, 1, 2],
    'b': [1, 0, 3],
    'c': [1, 1, 1]
})

schema = (
    FrameCheck()
    .columns(['a', 'b'], type='int', in_set=[0, 1])
)

result = schema.validate(df)

FrameCheck validation errors:
- Column 'a' contains values not in allowed set: [2].
- Column 'b' contains values not in allowed set: [3].

Go to Top

columns_are(...) – Exact column names and order

df = pd.DataFrame({'b': [1], 'a': [2]})

schema = FrameCheck().columns_are(['a', 'b'])
result = schema.validate(df)

FrameCheck validation errors:

Expected columns in order: ['a', 'b']

Found columns in order: ['b', 'a']

Go to Top

empty() – Ensure the DataFrame is empty

df = pd.DataFrame({'x': [1, 2]})

schema = FrameCheck().empty()
result = schema.validate(df)

FrameCheck validation errors:

DataFrame is expected to be empty but contains rows.

Go to Top

not_empty() – Ensure the DataFrame is not empty

df = pd.DataFrame(columns=['a', 'b'])

schema = FrameCheck().not_empty()
result = schema.validate(df)

FrameCheck validation errors:

DataFrame is unexpectedly empty.

Go to Top

only_defined_columns() – No extra/unexpected columns allowed

df = pd.DataFrame({'a': [1], 'b': [2], 'extra': [999]})

schema = (
FrameCheck()
.column('a')
.column('b')
.only_defined_columns()
)
result = schema.validate(df)

FrameCheck validation errors:

Unexpected columns in DataFrame: ['extra']

Go to Top

row_count(...) – Validate the number of rows

✅ Minimum rows

df = pd.DataFrame({'x': [1, 2]})

schema = FrameCheck().row_count(min=5)
result = schema.validate(df)

FrameCheck validation errors:

DataFrame must have at least 5 rows (found 2).

✅ Exact rows

df = pd.DataFrame({'x': [1, 2, 3]})

schema = FrameCheck().row_count(exact=2)
result = schema.validate(df)

FrameCheck validation errors:

DataFrame must have exactly 2 rows (found 3).

Go to Top

unique(...) – Rows must be unique

✅ All rows must be entirely unique

df = pd.DataFrame({
'user_id': [1, 2, 2],
'email': ['[e-mail]', '[e-mail]', '[e-mail]']
})

schema = FrameCheck().unique()
result = schema.validate(df)

FrameCheck validation errors:

Rows are not unique.

✅ Rows must be unique based on specific columns

df = pd.DataFrame({
'user_id': [1, 2, 2],
'email': ['[e-mail]', '[e-mail]', '[e-mail]']
})

schema = FrameCheck().unique(columns=['user_id'])
result = schema.validate(df)

FrameCheck validation errors:

Rows are not unique based on columns: ['user_id']

Go to Top

---

License

MIT

---

Contact

---

Go to Top