> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rowbase.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Data Quality

> Understand, correct, and validate your data

## Overview

Rowbase provides powerful tools to understand your data quality, make corrections, and ensure data meets your standards before use.

<CardGroup cols="2">
  <Card title="Diff View" icon="code-compare">
    Compare source data to transformed output
  </Card>

  <Card title="Inline Editing" icon="pen-to-square">
    Edit cells directly with automatic correction tracking
  </Card>

  <Card title="Data Profiling" icon="chart-bar">
    Get instant statistics for every column
  </Card>

  <Card title="Quarantine" icon="box-archive">
    Isolate problematic rows for review
  </Card>
</CardGroup>

## Diff View

Compare your source data to the transformed output to see exactly what changed.

### Using Diff View

1. Open a dataset with operations applied
2. Click the **Diff** button
3. View row-by-row changes

### What Diff Shows

| Status | Description |
| - | - |
| **Added** | New rows created by operations |
| **Removed** | Rows filtered out or quarantined |
| **Modified** | Rows with changed values |
| **Unchanged** | Rows that passed through unchanged |

The diff highlights:

* Which columns were affected
* Before and after values for each change
* Summary statistics (total added, removed, modified)

<Tip>
  Use diff view after applying operations to verify the transformations did what you expected.
</Tip>

## Data Profiling

Get instant statistics and distributions for every column.

### Profile Information

For each column, the profile shows:

* **Type** - Detected data type (text, number, date, boolean)
* **Unique values** - Count and percentage
* **Null values** - Count and percentage
* **Top values** - Most frequent values with counts

**For numeric columns:**

* Min, max, mean, median
* Standard deviation
* Histogram distribution

**For text columns:**

* Length statistics
* Character pattern analysis
* Common prefixes/suffixes

### Viewing the Profile

1. Open a dataset
2. Click the **Profile** tab
3. Explore statistics for each column

## Inline Cell Editing & Corrections

Edit any cell directly in the data grid. Your changes are tracked as corrections.

### How Corrections Work

Corrections are stored separately from your data and operations. They're applied last, after all operations run, so:

* Original data is never modified
* Corrections persist across new data versions
* You can see and manage all corrections

### Making a Correction

1. Click on a cell
2. Edit the value
3. The correction is saved automatically

### Viewing Corrections

Corrected cells show a visual indicator. You can:

* View all corrections in the **Corrections** panel
* See who made each correction and when
* Revert individual corrections

<Info>
  Corrections are applied via the **Apply Corrections** operation in your pipeline. This operation is added automatically when you make your first correction.
</Info>

## Quarantine

Isolate problematic rows for review without deleting them.

### Automatic Quarantine

Use the **Quarantine If** operation to automatically quarantine rows that fail validation:

```
Rule: {{price}} > 0
Result: Rows with zero or negative price are quarantined
```

### Manual Quarantine

You can also manually quarantine rows:

1. Select rows in the data view
2. Click **Quarantine**
3. Add a reason (optional)

### Working with Quarantined Rows

Quarantined rows are:

* Excluded from the main data view
* Stored in a separate quarantine table
* Available for review and restoration

To restore a quarantined row:

1. Open the **Quarantine** panel
2. Review the row and reason
3. Click **Restore** to bring it back

<Warning>
  Quarantined rows are excluded from exports and API responses by default.
</Warning>

## Validation Operations

Build validation rules into your pipeline:

| Operation | Purpose |
| - | - |
| **Require Fields** | Ensure columns are not null |
| **Unique Constraint** | Ensure values are unique |
| **Primary Key** | Combine uniqueness + not-null |
| **Validate Schema** | Check types match expectations |
| **Quarantine If** | Quarantine rows failing a formula |

### Quality Flags

Flag rows with potential issues without quarantining them:

* **Flag Negative** - Negative numeric values
* **Flag Zero** - Zero values
* **Flag If Less Than** - Column A \< Column B
* **Flag Invalid Units** - Units not in allowed list

Flags add a column indicating which rows have issues, so you can filter and review them.

## Best Practices

<AccordionGroup>
  <Accordion title="Profile before transforming" icon="chart-bar">
    Run the profiler first to understand your data. This helps you choose the right operations.
  </Accordion>

  <Accordion title="Use quarantine over delete" icon="box-archive">
    When in doubt, quarantine problematic rows instead of deleting. You can always restore them later.
  </Accordion>

  <Accordion title="Check the diff" icon="code-compare">
    After applying operations, use diff view to verify the changes match your expectations.
  </Accordion>

  <Accordion title="Build validation into pipelines" icon="shield-check">
    Add validation operations so future data imports are automatically checked against your rules.
  </Accordion>
</AccordionGroup>