diagnostics

Diagnostics

qrate checks project data continuously and lists what it finds in the Problems panel, in the bottom dock. Show or hide the panel with View ▸ Toggle Problems Panel or Ctrl+Shift+M. A finding on a cell also shows as a small marker on that cell in the grid.

What qrate checks

• Spelling, in over 60 languages. Tick one or more in Settings ▸ Spelling; each value is checked in whichever ticked language fits it, and changes apply without a restart; short or ambiguous values are left alone instead of being checked as the wrong language. If only English dictionaries are installed, qrate also skips reliably identified non-English text and common non-English apostrophe elisions.
• Capitalization, when a dictionary knows the word but requires a different case. Capitalization findings are warnings. The Notes tab is reserved for user notes.
• Value variants, for columns where you opt into reviewing inconsistent displayed forms. This catches punctuation, diacritic, word-order, and close-spelling differences in names, organizations, places, subjects, titles, and other labels. Similarity is a review hint, not a claim that two values identify the same entity. Right-click a column header and select Review value variants to enable this check. Select the checked command again to disable it.
• Date formats, so a malformed or ambiguous date is caught before export. Choose EDTF, ISO 8601 only, or Lenient in Settings ▸ Table ▸ Checks ▸ Date format; see Columns.
• File links, so a row whose linked file cannot be found is reported instead of silently showing a blank preview. See Files and photos. Only a column set to the Filename type is checked this way.
• Headings against LCSH, GeoNames, and Wikidata, so a subject or place heading can be checked against those authorities.
• Plugin validators, if the project has plugins that add their own checks. See Plugins.

Each of these runs as an independent validator. A validator reports its complete set of findings each time it runs, and that set replaces what it reported last time.

The Problems panel

Repeated spelling, capitalization, value-variant, date, and authority findings appear in collapsed groups. Each group shows an occurrence count. Expand the group to see its locations. Expanded value-variant groups show the value at each location instead of repeating the summary. The expansion button supports keyboard activation. Click an occurrence to select its cell, row, or column. A group header does not select an arbitrary location. Its context menu can resolve all occurrences.

Findings can describe four scopes:

Scope	Location label	Navigation
Cell	Row N · Column	Select the cell
Row	Row N	Select the row
Column	Column name	Select the column
Dataset	Dataset name	No cell selection

The severity tabs count diagnostic occurrences, not collapsed groups or unique cells. Two distinct misspelled words in one cell count as two occurrences. Repeated copies of the same word in one cell count once. A cell can contain more than one member of a value-variant cluster.

The source filter changes both the visible findings and the tab counts. It is a multi-select filter. Uncheck one or more diagnostic sources to hide them. User notes do not appear as a source because the Notes tab already selects them. The All, Errors, and Warnings tabs show computed findings only. The Notes tab shows user notes only. The severity filter changes the visible findings, but each tab keeps its own total. Notes and validators without group metadata remain separate entries.

Producer contracts

The diagnostics store keeps each finding at its exact location. The panel groups findings only when a producer supplies DiagnosticGroup metadata. Group identity includes the dataset, source, severity, and producer key. The summary is display text, not identity.

Column validators return ColumnFinding. A row index describes a cell. An absent row index describes the whole column. Row-wide and dataset-wide producers publish addressed diagnostics directly. Existing plugin scripts keep their current row-based output format.

Spelling keys include the selected dictionary and observed token. Capitalization keys also include the proposed spelling. Value-variant keys include the column and the sorted set of exact displayed cluster members. Similarity pairs form links. Connected values appear together for review, but qrate does not change them until the user chooses a canonical form. Date keys are the exact rejected value. Authority keys are the rejected value, ignoring case.

Applying a fix

Some findings offer a suggested correction. Right-click the cell, open Problems, choose the finding, and pick a suggestion. The same items appear when you right-click the finding in the Problems panel. Long labels are shortened, and a cell with more than eight findings lists the first eight and points to the Problems panel for the rest. Applying a fix replaces the cell's whole value, so you always know exactly what you are accepting.

Value-variant fixes offer forms that already occur in the same column. qrate never chooses a canonical form, merges records, or changes every matching row automatically.

Right-click a spelling or capitalization group to apply one correction to all its occurrences. The group resolver applies the changes as one undo step.

A misspelled word's menu also offers Add “word” to dictionary. qrate stops flagging that word in every project, and keeps it in dictionary.txt in qrate's data folder, so it stays accepted after a restart.

Right-click a value-variant group to replace all occurrences with any displayed member. You can also mark the cluster members as distinct for that column. qrate saves this choice with the project.

qrate uses the configured subdelimiter to read multiple logical values in one cell. Validators borrow these values as text slices, so splitting does not allocate a list for each cell. Value-variant fixes preserve the other logical values and replace the complete cell through the normal edit path.

A fix offered against one version of a cell's text does not apply once that text has changed. This stops a stale suggestion from silently overwriting a newer edit.

Expanded cell occurrences offer the same spelling and fix menus as the grid. Row, column, dataset, and group entries do not offer cell fixes. Each accepted fix uses the existing undoable cell edit path. Validation then refreshes the groups.

Ignoring a finding

Each finding offers two ways to ignore it:

• Ignore this occurrence hides it in this one cell. qrate remembers the row, so the ignore follows the row when you insert or delete rows above it.
• Ignore in column hides every finding with the same key in that column. For spelling, the key includes the word, so a different misspelling still shows.

A group row in the Problems panel offers Ignore ▸ In column for each column it covers. After you ignore something, a notice with Undo appears for a few seconds. qrate saves ignores with the project. They never hide findings from another check.

Turn on Show ignored in the Problems panel to see ignored findings dimmed, and right-click one to Unignore it. Settings ▸ Columns ▸ Ignored problems lists every ignore that still hides something. Remove one there, or use Clear all.

Design credit

qrate's value-clustering workflow is inspired by OpenRefine's Cluster and edit feature and clustering methods. qrate uses an independent implementation for its diagnostics and cataloging workflow. This acknowledgment does not imply OpenRefine's endorsement.

OpenRefine publishes its source under BSD-3-Clause and its documentation under CC BY 4.0. This implementation does not copy OpenRefine source, documentation text, or UI assets. Any future adaptation must retain the applicable copyright, license, and attribution notices.

---

Generated from the source documentation. Edit that file to update this page.