dataexaminer

A Python library that examines datasets and produces per-column summaries including data type, record count, missing values, and anomaly detection.

Installation

pip install dataexaminer

For Excel file support:

pip install dataexaminer[excel]

Quick Start

from dataexaminer import examine

# From a CSV or Excel file
result = examine("data.csv")             # returns a dict
examine("data.csv", output="console")   # prints a table

# From a pandas DataFrame
import pandas as pd
df = pd.read_csv("data.csv")
examine(df, output="console")

Example Output

DataExaminer Report — sales_data.csv
Examined: 2026-03-22 14:05:00   Rows: 5,000   Cols: 4

Column      Dtype           Kind         Records  Missing  Miss%   Anomalies  Detail
----------  --------------  -----------  -------  -------  ------  ---------  -------------------------------------------------
revenue     float64         numeric      5,000    23       0.5%    12         12 IQR outliers (fence: 105.0 – 9842.0)
category    object          categorical  5,000    41       0.8%    8          8 values in rare categories (<1%): ['UNKN', 'N/a']
order_date  datetime64[us]  datetime     5,000    5        0.1%    3          3 gap(s) >5.0× median (median: 1d, largest: 47d)
notes       object          categorical  5,000    5,000    100.0%  0          —

What Gets Detected

Programmatic Usage

from dataexaminer import DataExaminer
from dataexaminer.formatters import ConsoleFormatter, DictFormatter

examiner = DataExaminer()
report = examiner.examine(df)

# report.columns is a list of ColumnSummary objects
for col in report.columns:
    print(col.column_name, col.missing_count, col.anomaly_count)

Working with the Report Object

report = examiner.examine(df)

print(report.row_count)       # total rows
print(report.col_count)       # total columns
print(report.examined_at)     # datetime of analysis

for col in report.columns:
    print(col.column_name)    # column name
    print(col.dtype)          # pandas dtype string
    print(col.kind)           # "numeric" | "categorical" | "datetime" | "boolean"
    print(col.record_count)   # total rows
    print(col.missing_count)  # number of null values
    print(col.missing_pct)    # percentage of null values
    print(col.anomaly_count)  # number of anomalies detected
    print(col.anomaly_detail) # human-readable description

Dict Output

result = examine("data.csv")  # default output="dict"
print(result["row_count"])
print(result["columns"][0]["anomaly_detail"])

Custom Detectors

The library is designed to be extended. Implement AnomalyDetector and pass it in:

from dataexaminer import DataExaminer
from dataexaminer.detectors import AnomalyDetector, IQRAnomalyDetector
import pandas as pd

class ZeroInflationDetector(AnomalyDetector):
    """Flags columns where more than 50% of non-null values are zero."""

    def supports(self, series: pd.Series) -> bool:
        return pd.api.types.is_numeric_dtype(series)

    def detect(self, series: pd.Series) -> tuple[int, str]:
        clean = series.dropna()
        if clean.empty:
            return 0, "—"
        zeros = (clean == 0).sum()
        pct = zeros / len(clean) * 100
        if pct > 50:
            return int(zeros), f"{zeros} zero values ({pct:.1f}%)"
        return 0, "—"

examiner = DataExaminer(detectors=[IQRAnomalyDetector(), ZeroInflationDetector()])
report = examiner.examine(df)

Tuning Built-in Detectors

from dataexaminer.detectors import IQRAnomalyDetector, RareCategoryDetector, DatetimeAnomalyDetector

examiner = DataExaminer(detectors=[
    IQRAnomalyDetector(iqr_multiplier=3.0),       # less sensitive to outliers
    RareCategoryDetector(threshold_pct=5.0),       # flag categories below 5%
    DatetimeAnomalyDetector(gap_multiplier=10.0),  # only flag very large gaps
])

Supported File Formats

Requirements

• Python 3.10+
• pandas 2.0+

License

MIT © Aeron Zentner