intake

Phase 1 — data intake summary (shape, dtypes, missingness, descriptives).

intake

Phase 1 — inspect the dataset as loaded and surface its shape and contents.

Reports the shape, dtypes, missing-value counts, numeric descriptives, and categorical frequencies of the dataset as loaded, so the researcher can confirm the file the harness sees matches their raw file.

inspect(df)

Summarize a dataset's shape, dtypes, missingness, and value distributions.

Builds the Phase-1 intake record: informational CheckResult lines for the dataset dimensions, per-column dtypes, and missing-value counts, plus rendered artifacts (a 10-row head, numeric describe() output, and top-10 categorical frequency tables) for the verification log. All entries are informational; nothing here passes or fails.

Parameters:

Name	Type	Description	Default
df		The dataset as loaded, inspected exactly as received without modification.	required

Returns:

Type	Description
	A (results, artifacts) tuple, where results is a list of informational
	CheckResult records and artifacts is a dict mapping "head",
	"describe", and "categorical" to pre-rendered string blocks.

Source code in crossverify/intake.py

def inspect(df):
    """Summarize a dataset's shape, dtypes, missingness, and value distributions.

    Builds the Phase-1 intake record: informational ``CheckResult`` lines for the
    dataset dimensions, per-column dtypes, and missing-value counts, plus rendered
    artifacts (a 10-row head, numeric ``describe()`` output, and top-10 categorical
    frequency tables) for the verification log. All entries are informational; nothing
    here passes or fails.

    Args:
        df: The dataset as loaded, inspected exactly as received without modification.

    Returns:
        A ``(results, artifacts)`` tuple, where ``results`` is a list of informational
        ``CheckResult`` records and ``artifacts`` is a dict mapping ``"head"``,
        ``"describe"``, and ``"categorical"`` to pre-rendered string blocks.
    """
    results = []
    artifacts = {}

    results.append(
        CheckResult(
            "1",
            "intake:shape",
            "Dataset dimensions",
            None,
            f"{df.shape[0]} rows x {df.shape[1]} columns",
        )
    )

    dtypes = "; ".join(f"{c}: {t}" for c, t in df.dtypes.astype(str).items())
    results.append(CheckResult("1", "intake:dtypes", "Column dtypes", None, dtypes))

    miss = {c: int(n) for c, n in df.isna().sum().items() if n > 0}
    detail = "none" if not miss else "; ".join(f"{c}: {n}" for c, n in miss.items())
    results.append(CheckResult("1", "intake:missing", "Missing values per column", None, detail))

    artifacts["head"] = df.head(10).to_string()
    numeric = df.select_dtypes("number")
    artifacts["describe"] = (
        numeric.describe().to_string() if not numeric.empty else "(no numeric columns)"
    )

    cat_blocks = []
    for col in df.select_dtypes(exclude="number").columns:
        vc = df[col].value_counts().head(10)
        cat_blocks.append(
            f"{col} (top {min(10, df[col].nunique())} of {df[col].nunique()} unique):\n{vc.to_string()}"
        )
    artifacts["categorical"] = "\n\n".join(cat_blocks) if cat_blocks else "(no categorical columns)"

    return results, artifacts

numeric_ranges(df)

Compute the (min, max) range of each numeric column.

Used by centroid-in-range consistency checks, which confirm a reported centroid falls within the observed span of its column.

Parameters:

Name	Type	Description	Default
df		The frame to scan; only number-dtype columns are included.	required

Returns:

Type	Description
	A dict mapping each numeric column name to a (min, max) tuple of float.

Source code in crossverify/intake.py

def numeric_ranges(df):
    """Compute the (min, max) range of each numeric column.

    Used by centroid-in-range consistency checks, which confirm a reported centroid
    falls within the observed span of its column.

    Args:
        df: The frame to scan; only number-dtype columns are included.

    Returns:
        A dict mapping each numeric column name to a ``(min, max)`` tuple of ``float``.
    """
    numeric = df.select_dtypes("number")
    return {c: (float(numeric[c].min()), float(numeric[c].max())) for c in numeric.columns}