Comparison.get_tabular_report()

Get a tabular report of the comparison.

Usage

Comparison.get_tabular_report(
    limit=10,
    title=None,
    lang=None,
    locale=None,
)

The report brings everything about a comparison together in one table, for reading rather than for further processing. It’s what’s shown when a comparison is displayed in a notebook, and calling get_tabular_report() directly lets the report be customized and saved, for example to attach to a pull request or to share with whoever owns the data.

The report has a header with the verdict, the row status counts, and the comparison settings, followed by sections (as row groups) for schema differences, per-column mismatches, and samples of changed, missing, extra, and duplicate-key rows. Changed rows show only the columns that differ, with the source and target values (and the difference, for numbers). Sections with nothing to show are omitted.

Parameters

limit: int = 10: The maximum number of sample rows shown per row status.
title: str | None = None: An optional title replacing the default.
lang: str | None = None: The language of the report’s labels, as a two-letter language code (e.g., "fr", "de") or one of the other Pointblank reporting languages. English by default. Values from the tables (keys, column names, and data) are shown as they are.
locale: str | None = None: The locale used to format counts and percentages (e.g., the thousands separator). By default, it’s the same as lang=.

Returns

GT: A Great Tables object.

Examples

Let’s compare two months of a payroll table. Naming the tables with source_name= and target_name= makes the report easier to read.

import pointblank as pb
import polars as pl

payroll_jan = pl.DataFrame(
    {
        "emp_id": [101, 102, 103, 104],
        "dept": ["Sales", "Sales", "Ops", "Ops"],
        "salary": [52000, 61000, 58000, 47000],
    }
)
payroll_feb = pl.DataFrame(
    {
        "emp_id": [101, 102, 103, 105],
        "dept": ["Sales", "Sales", "Ops", "Ops"],
        "salary": [52000, 63500, 58000, 49000],
    }
)

comparison = pb.compare(
    payroll_jan,
    payroll_feb,
    keys="emp_id",
    source_name="payroll_jan",
    target_name="payroll_feb",
)

comparison.get_tabular_report()

Column mismatches (1 of 2 columns)
Table Comparison
payroll_jan (4 rows) → payroll_feb (4 rows), aligned by key `emp_id` DIFFERENT 3 of 5 rows fail (60.0%) match 2changed 1missing 1extra 1
salary	1 of 3 matched rows differ (33.3%)
Changed rows (1 row)
102	salary: 61000 → 63500 (+2500)
Missing from target (1 row)
104	dept: Ops salary: 47000
Extra in target (1 row)
105	dept: Ops salary: 49000

The verdict and counts are at the top, and below them are the sections that have something to show: the mismatches in salary, the changed row for employee 102, and the employees who left and joined.

The title can be replaced, and the labels can be shown in any of Pointblank’s reporting languages. Here’s the same report in French:

comparison.get_tabular_report(title="Paie : janvier et février", lang="fr")

Écarts par colonne (1 colonnes sur 2)
Paie : janvier et février
payroll_jan (4 lignes) → payroll_feb (4 lignes), alignées par la clé `emp_id` DIFFÉRENTES 3 lignes sur 5 en échec (60,0%) identique 2modifiée 1manquante 1supplémentaire 1
salary	1 lignes appariées sur 3 diffèrent (33,3%)
Lignes modifiées (1 ligne)
102	salary: 61000 → 63500 (+2500)
Absentes de la cible (1 ligne)
104	dept: Ops salary: 47000
Supplémentaires dans la cible (1 ligne)
105	dept: Ops salary: 49000

Numbers are formatted for the French locale too, with a decimal comma (as in 60,0%). Use locale= to format them differently from the labels. To save a report as a standalone HTML file, use the write_raw_html() method of the returned Great Tables object.