{{ source_stem }}{{ source_suffix }}

Rows

{{ ds.summary.row_count | count }}{% if ds.summary.empty_row_count %}{{ ds.summary.empty_row_count | count }} empty rows{% else %}No empty rows{% endif %}

Columns

{{ ds.summary.column_count | count }}{{ ds.summary.cell_count | count }} cells analyzed

Missing cells

{{ ds.summary.missing_percent }}%{{ ds.summary.missing_count | count }} cells

Duplicate rows

{% if ds.summary.duplicate_row_status == 'disabled' %}–Not checked{% elif ds.summary.duplicate_row_status == 'limited' %}≥ {{ ds.summary.duplicate_row_count | count }}At least, beyond the first occurrence{% else %}{{ ds.summary.duplicate_row_count | count }}Beyond the first occurrence{% endif %}

Not computed: the scan has no observations (scan.observations.enabled is false, or building them failed).

No observations: the scan found nothing to report.

Confirmed issues

{{ qc.confirmed_issue | count }}Defects of the source

Candidate anomalies

{{ qc.candidate_anomaly | count }}Worth a look

Remarks

{{ qc.informational | count }}May be valid

Limits

{{ ds.quality.summary.limits | count if ds.quality.summary else '–' }}What the scan could not establish

Showing the first {{ display_limits.findings | count }} of {{ ds.quality.items | length | count }} findings; the report JSON lists them all.

The scan listed only its first scan.observations.max_items findings; its summaries count all of them.

Accuracy: exact counted on every value of the file; at least a scan limit stopped the count; sample counted on a sample of the distinct values. Remarks such as missing values are not quality issues.

Distinct values are exact unless marked: ≥ at least (a scan limit stopped the count), ≈ estimated from a sample with its error.

Records

{{ ds.summary.row_count | count }}{% for name, count in st.record_types.items() %}{{ count | count }} {{ name }}{% if not loop.last %}, {% endif %}{% else %}No records{% endfor %}

Paths

{% if st.path_count is none %}limited{% else %}{{ st.path_count | count }}{% endif %}{{ ds.summary.column_count | count }} with values

Maximum depth

{{ st.max_depth_seen }}Limit {{ report.config.limits.max_depth }}

No paths: the dataset has no records.

Distribution over time exact

Valid dates counted by the second pass per calendar period, as written: the offset of a datetime is not converted, and ambiguous dates are counted apart and never placed, unless the unambiguous dates of the column prove one order and the second pass reads them with it.

Columns with patterns

{{ pattern_columns | length | count }}of {{ ds.summary.column_count | count }} columns

Without patterns

{{ pattern_missing | count }}Numbers, dates, masked or too few distinct values

Grammar

{{ pattern_grammar }}Version of the pattern rules

Showing the first {{ display_limits.patterns | count }} of {{ pattern_rows | length | count }} patterns; the report JSON lists them all.

A share is the part of the distinct values a pattern takes: exact when the scan held every distinct value, sample when it held a sample of them. A value counts for one pattern: the one of its family that has its lengths. Quoted text is a constant the pattern folds. Validated on every value is the second pass: how many occurrences of the whole file the pattern reads, exactly; a value no pattern reads is an observation of structure, not an error.

Showing the first {{ display_limits.relations | count }} of {{ rel.groups | length | count }} relations; the JSON lists them all.

{{ note }}

Correlations between numeric columns({{ rel.correlations | length }})

Counts are exact: every row of the file is placed in one category and one class, or counted as left out. The dominant class of a category is the one that holds most of it; a tie goes to the first class listed. Purity is the share of the classified rows that sit in the dominant class of their category. The score is Cramér's V corrected for the number of rows and classes; when the expected counts are low it is marked unstable and given no strength. An empty other row or column is not printed. Correlations show a linear association between two columns, over the rows where both have a number.

{{ family_titles[card.family] }}

Showing {{ v.insight_cards | length | count }} of {{ ds.insights | length | count }} cards, at most {{ display_limits.insights | count }} of each family; the report JSON holds every result.

{{ sp.reason }}

Computed

{{ m.computed | count }}Analyses with a result

Not computed

{{ (m.skipped + m.failed) | count }}{{ m.skipped | count }} skipped, {{ m.failed | count }} failed

Already answered

{{ m.already_answered | count }}By the scan, no reading needed

Not selected

{{ m.not_selected | count }}Left out by the planner

No analysis was run for this dataset.

Not selected, by reason: {% for reason, count in sp.left_out %}{{ count | count }} {{ reason }}{% if not loop.last %}; {% endif %}{% endfor %}.

The source was read again{% if sp.source_verified %} and checked against the digest of the scan, so no result mixes two versions of the data{% endif %}{% if sp.seconds is not none %}, {{ sp.seconds | seconds }} s{% endif %}. This run computed {{ sp.executed | count }} analyses and reused {{ sp.reused | count }} stored ones.

The source was not read for this report. Stored results and the facts of the scan were used.

Untracked observations

{{ lim.untracked_observations | count }}Paths beyond scan.limits.max_fields

Truncated by depth

{{ lim.depth_truncated_observations | count }}Deeper than scan.limits.max_depth

The limited measures are not computed: the scan has no observations (scan.observations.enabled is false, or building them failed).

Limited measures({{ lim.items | length }})

Showing the first {{ display_limits.limits | count }} of {{ lim.items | length | count }} limits; the report JSON lists them all.

Diagnostics({{ lim.diagnostics | length }})

{% for d in lim.diagnostics %}
• {{ d.level }} {{ d.code }} {{ d.message }}{% if d.dataset is none %} whole scan{% endif %}{% if d.locations %}Records {% for loc in d.locations %}{{ loc.record }}{% if loc.line is not none %} (line {{ loc.line }}){% endif %}{% if not loop.last %}, {% endif %}{% endfor %}{% if d.count > d.locations | length %}…{% endif %}{% endif %}{{ d.count | count }}
{% endfor %}

Scope

All {{ ds.summary.row_count | count }} records

{% set scan = report.config %}

Missing values

{{ scan.values.missing | join(', ') }}; markers: {% for marker in scan.values.null_markers %}{{ marker | tojson }} {% else %}none{% endfor %}

Marker comparison

Whitespace trimmed; {{ 'case-sensitive' if scan.values.null_markers_case_sensitive else 'case-insensitive' }}

Value normalization

Unicode composition (NFC): {{ 'enabled' if scan.normalization.nfc else 'disabled' }}; trim: {{ 'enabled' if scan.normalization.trim else 'disabled' }}; collapse internal whitespace: {{ 'enabled' if scan.normalization.collapse_whitespace else 'disabled' }}; comparison also with case folding: {{ 'enabled' if scan.normalization.casefold else 'disabled' }}, accents removed: {{ 'enabled' if scan.normalization.strip_accents else 'disabled' }}. Raw preview values are preserved.

Duplicate comparison

{% if scan.records.duplicates %}Exact raw values across every column; up to {{ scan.limits.max_tracked_records | count }} distinct records compared{% else %}Disabled{% endif %}

Type inference

At least {{ (scan.types.minimum_confidence * 100) | round(1) }}% agreement across present values; numbers with {{ scan.detectors.number.conventions | join(' or ') }} decimals, configured dates and true/false. Leading-zero identifiers remain text.

Date detection

{% if scan.detectors.date.enabled %}Orders: {{ scan.detectors.date.orders | join(', ') }}; separators: {% for separator in scan.detectors.date.separators %}{{ separator | tojson }} {% endfor %}; month names: {{ scan.detectors.date.month_languages | join(', ') or 'none' }}; ambiguous order: {{ scan.detectors.date.ambiguous_order or 'unresolved, column evidence shown but never applied' }}{% else %}Disabled{% endif %}

Column types

The technical type, a detector matching at least {{ (scan.detection.minimum_share * 100) | round(1) }}% of present values, the length of the texts, the name of the column and, when the second pass counted them, the exact distinct values

Adaptive detection

{% if scan.detection.warmup_values %}Every detector on the first {{ scan.detection.warmup_values | count }} distinct values of each column; detectors that recognized none of them{% if scan.detection.rare_share %}, or at most {{ (scan.detection.rare_share * 100) | round(2) }}% of them and none of the last {{ (scan.detection.warmup_values - scan.detection.warmup_values // 2) | count }},{% endif %} then skip the others{% if scan.detection.probe_interval %}, except about 1 value in {{ scan.detection.probe_interval | count }}{% endif %}. Number and date detection always test every value{% else %}Disabled: every detector tests every value{% endif %}

Enumerations

{% if scan.detectors.enumeration.enabled %}At most {{ scan.detectors.enumeration.maximum_distinct }} distinct values among at least {{ scan.detectors.enumeration.minimum_values | count }} present values{% else %}Disabled{% endif %}

Sensitive values

{{ {'mask': 'Masked', 'hide': 'Hidden', 'show': 'Shown'}[scan.exposure.sensitive_values] }} in examples and the data sample

String lengths

Very short through {{ report.config.classification.very_short_max_length }}; short through {{ report.config.classification.short_max_length }}; medium through {{ report.config.classification.medium_max_length }}; long through {{ report.config.classification.long_max_length }}; otherwise very long. Up to {{ report.report.config.string_analysis.examples_per_length }} examples per length are retained when the maximum is {{ report.report.config.string_analysis.length_distribution_max_length }} or less.

Value representation

Complete through {{ report.report.config.value_examples.full_distribution_max_distinct }} distinct values; otherwise a reproducible sample of up to {{ v.sample_size }} distinct values, the ones with the smallest hash priorities

Row numbering

Data records start at 1, excluding the header. Quoted multiline values count as one record.

Row sample

{% if ds.rows.sample %}The scan kept the first {{ ds.rows.sample.requested.edge }} records, up to {{ ds.rows.sample.requested.middle }} records drawn from the interior by a seeded hash, and the last {{ ds.rows.sample.requested.edge }}; this report shows {{ report.report.config.row_sample.start }} at the start, {{ report.report.config.row_sample.middle }} sampled and {{ report.report.config.row_sample.end }} at the end; raw values preserved{% else %}Disabled{% endif %}

Patterns

{% if scan.pattern_inference.enabled %}Inferred from the sampled distinct values: up to {{ scan.pattern_inference.max_patterns }} per column, values up to {{ scan.pattern_inference.max_value_length }} characters, at least {{ scan.pattern_inference.min_distinct }} distinct values per pattern{% else %}Not computed: disabled by scan.pattern_inference.enabled{% endif %}

Observations

{% if scan.observations.enabled %}Rules version {{ report.observations.version if report.observations else '–' }}; at most {{ scan.observations.max_items | count }} findings and {{ scan.observations.max_items | count }} limits listed{% else %}Not computed: disabled by scan.observations.enabled{% endif %}

JSON format

{{ report.format_version }} - revision {{ report.format_revision }} (experimental)