Table Explorer

Table Explorer provides a powerful and readily accessible data exploration tool with built in filtering, summarization, and other features to make life easy for people working with large and complex data.

Table Explorer supports exploration on any size dataset so you can use the same tool no matter how much your data grows. It also provides point-and-click filtering along with advanced filter capabilities to zero in on the data you need. The best part is that anywhere in PlaidCloud with tables or views, you can click on those tables and views to explore with Table Explorer. By being fully integrated, data access is only a click away.

The Grid view provides a tabular view of the data. The Details view provides a summary of each column, a count of unique values, and profiling statistics for every column.

Data can be exported directly from a filtered set as well as being able to save and share filters with others. Finally, the filters and column settings can be saved directly as a workflow Extract step.

The Grid View

The Grid view provides a tabular view of the data.

Setting the Row Limit

By default, the row limit is set to 5,000 rows. However, this can be adjusted or disabled entirely.

The rows shown along with the total size of the dataset are shown at the bottom of the table. The information provides three key pieces of information:

1. The current row count shown based on the row limit applied
2. The size of the global data after filters are applied
3. The size of the unfiltered global data

Caution

Be careful not to disable the row limit functionality when viewing larger (e.g. millions of rows) because this could cause your browser to run slow. Try using filters to find the data instead.

Choosing Top or Random Rows

Next to the row limit is a Top / Random selector that controls which rows fill the grid:

• Top (the default) returns the first rows of the table — the fastest option, and the right choice when you just want a quick look.
• Random returns a fast random sample drawn from across the whole table, so the preview reflects the full range of your data rather than just the beginning. This is useful for spot-checking data quality on a large table where the first rows aren’t representative.

Random sampling uses your storage engine’s built-in sampling, so it stays quick even on very large tables. It returns the number of rows you asked for — drawn at random from across the table, or all of them if the table holds fewer rows than your limit — and you get a different sample each time you refresh. The selector applies to the grid preview only; it has no effect on saved views, extracts, downloads, or the reported row counts. Random requires the row limit to be enabled.

Note

Sorting by a column header sorts only the rows currently retrieved, so a sort combined with Random orders just the sampled rows, not the whole table.

Sorting Locally Versus Globally

The Grid view provides the ability to click on the column header and sort the data based on that column. However, this method is only sorting the dataset that has already been retrieved and is not sorting based on the full dataset. If your retrieved data contains the entire dataset this distinction is immaterial however if your full dataset is larger than what appears in the browser, this may not be the desired sort result.

If you desire to sort the global dataset before retrieving the limited data that will appear in your browser those sorts can be applied to the columns in the Details view by clicking on the Sort icon at the top of each column. An additional benefit of using the global sort approach is that you can apply multiple sorts along with a mix of sort directions.

Quick Reference Column List

All of the columns in the table or view are shown on the left of the Table Explorer window by default. This column list can be toggled on and off using the column list toggle button.

The column list provides a number of quick access and useful features including:

• Double clicking an item jumps to the column in the Grid or Details view
• Control visibility of the column through the visibility checkbox
• Use multi-select and right-click to include or exclude many columns at once
• Quickly view the data type of each column using the data type icons
• View the total column count

The Details View

The Details view provides an efficient way to view the data at a high level and exposes tools to quickly filter down to information with point-and-click operations.

Note

Column summaries are not automatically generated for views. You can click on the column refresh button to calculated the details though.

Column Data and Unique Counts

Each column is shown, provided it is currently marked as visible. The column summary displays the top 1,000 unique values by count. The number of unique values shown can be adjusted by selecting the Detailed Rows Displayed selection for a different value.

Managing Point-and-Click Filters

Each column provides for point-and-click filtering by activating the filter toggle at the top of the column. Select the items in the column that you would like to include in the resulting data. Multi-select is supported.

Once you apply a filter, there may be items you wish to remove or to clear the entire column filter without clearing all filters. This is accomplished by selecting the dropdown on the column filter button and unchecking columns or selecting the clear all option at the top.

Managing Summarization

Summarization of the data can be applied by toggling the Summarize button to On. When the Summarize button is activated, each column will display a summarization type to apply. Adjust the summarization type desired for each column.

When the desired summarizations are complete, refresh the data and the summarizations will be applied.

Examples of summarization types are Min, Max, Sum, Count, and Count Distinct.

Finding Distinct Values

Activating the Distinct button will help reduce the data to only a set of unique records. When the Distinct button is active, a Distinct checkbox will appear on each column. Uncheck the columns that DO NOT define uniqueness of the column to the dataset. For example, if you want to find the unique set of customers in a customer order table, you would only want to select the customer column rather than including the customer order number too.

Caution

If you include too many columns in the unique records determination, it will appear you have many more distinct results than you should.

Column Profiling Statistics

Every column displays profiling statistics at the bottom of its column information, regardless of data type.

Integer and numeric columns display these statistics, calculated on the full filtered dataset:

• Minimum
• Maximum
• Mean
• Std. Deviation
• Sum

Every column — numeric, text, date, or otherwise — also displays these statistics, calculated on the full unfiltered table:

• Null Count (All Rows) — the number of null values across the whole table
• Null % (All Rows) — null values as a percentage of the whole table
• Blank Count (All Rows) — text columns only. A blank is a value that’s empty or whitespace-only after trimming, as distinct from a null.
• Blank % (All Rows) — blank values as a percentage of the whole table
• Distinct % of Non-Null (All Rows) — distinct values as a percentage of the rows that actually have a value. Nulls are excluded from both the distinct count and the total, so a column that’s 90% null but whose populated values are all unique reads 100%, not 10%.

Note

The (All Rows) suffix is there for a reason: those five statistics always cover the whole table, ignoring any filter you’ve applied — unlike Minimum, Maximum, Mean, Std. Deviation, and Sum above, which respect your current filter. Two differently-scoped sets of numbers in one panel need the label to avoid being misread as the same scope.

These (All Rows) statistics are computed server-side and deterministically, with no AI or sampling involved — the same column returns the same numbers every time. They respect row-level security, so you only ever see statistics computed over rows you’re entitled to see. Computing them costs one additional whole-table query, subject to the same 60-second budget as other column facets; if that query times out, the rest of the grid is unaffected.

Distribution Shape

For numeric columns, Table Explorer also reports how the values are distributed, not just where they cluster.

• Skewness (All Rows) — measures whether the column’s values lean to one side of the average or are evenly balanced around it. A value near zero means the values are roughly symmetric; a strongly positive or negative value means the column has a long tail stretching toward one end. This can reveal a shape that Mean and Std. Deviation alone would hide — for example, a handful of very large values pulling the average up while most of the column sits much lower.
• A histogram below each numeric column’s statistics groups its values into 10 equal-width buckets spanning the column’s full range, so you can see the shape of the distribution at a glance. Hover over a bar to see the exact value range it covers and how many rows fall in it.

Like the other (All Rows) statistics, Skewness and the histogram always cover the whole table regardless of your current filter. Skewness is omitted — never shown as zero or an error — for a column with fewer than two values or with no variation at all, since it can’t be calculated meaningfully in that case. The histogram is omitted the same way for a column with no computable range, such as one that’s entirely null.

Data Quality Radar

Table Explorer runs five automated checks against every column to flag data that’s worth a closer look:

• Outliers — numeric values statistically far from the rest of the column (more than three standard deviations from the mean).
• Format — text values that don’t match the dominant pattern found in the rest of the column, such as a column that’s almost entirely dates with a few free-text values mixed in. This check only runs on columns with at least 100 non-blank values, and only flags a mismatch once one pattern clearly dominates — more than 95% of the column’s values matching it.
• Duplicates — repeated values in a column that’s part of the table’s declared row key, where every value should be unique. This check only runs on tables with a declared row key — the table’s update keys are used as the row key.
• Orphans — values in a column linked to a dimension that don’t match any member of that dimension.
• Constant — every non-null value in the column is the same, which usually means the column carries no useful information.

These checks run over the whole table, the same as the (All Rows) statistics above — your current filter has no effect on them.

A badge appears directly on a column in the Details view for any check that found something, and a muted badge appears for a check that couldn’t complete. Open Data Quality Radar from the Actions menu for a ranked summary of every column against all five checks — including checks that came back clean or don’t apply, not just the ones that were flagged.

Each check independently reports one of five outcomes: flagged, clean, not applicable, timed out, or unavailable. A check is never silently skipped — you can always see why a column has no verdict for a given check.

Correlations

For numeric columns, Table Explorer also reports how strongly a column’s values move together with the table’s other numeric columns.

• Pearson r vs <column> (Complete Cases) — how strongly two columns move together in a straight-line relationship, from -1 (perfectly opposite) through 0 (no linear relationship) to +1 (perfectly together).
• Spearman ρ vs <column> (Complete Cases) — how strongly two columns move together in rank order, whether or not that relationship is a straight line. A pair that consistently rises together but curves rather than forming a straight line can score higher on Spearman than on Pearson.
• Complete Cases (All Rows) — how many rows the coefficients above it are actually computed from.

Correlations are computed against up to 10 other numeric columns currently visible in the Details view, and only using rows that have a value in every one of those numeric columns — not just the pair being compared. That’s why the scope is labelled (Complete Cases) rather than (All Rows): it’s a narrower, listwise subset of the table, and calling it (All Rows) would overstate what the coefficients describe. The Complete Cases (All Rows) count itself is measured against the whole table, the same as the other (All Rows) statistics — it isn’t affected by your current filter, though it does change if you change which numeric columns are visible.

Correlations are omitted entirely, rather than shown from too little evidence, unless there are at least 10 complete-case rows and they make up at least 10% of the table. A single partner column whose own coefficient can’t be computed — no variation in that column, for example — drops just that one row rather than the whole set.

Copying Data

It is sometimes useful to allow for copying of selected data from PlaidCloud so that it can be pasted into other applications such as a spreadsheet.

From the Copy button in the upper right, there are several copy options available for the data:

• Copy All - Copies all of the data to the clipboard
• Copy Selection - Copies the selected data to the clipboard
• Copy Cell - Copies only the contents of a single cell to the clipboard
• Copy Column - Copies the full contents of the column to the clipboard

Exporting Data

Exporting data from the Table Explorer interface allows exporting of the filtered data with only the columns visible. You can export in the following formats:

• Microsoft Excel (xlsx)
• CSV (Comma)
• TSV (Tab)
• PSV (Pipe)

The Download menu also offers the ability to download only the rows visible in the browser. This is based on using the row limit specified.

Additional Actions

Additional useful actions are available under the Actions menu.

Save as Extract Step

When exploring data, it is often in the context of determining how to filter it for a data pipeline process. This often consists of applying multiple filters including advanced filters to zero in on the desired result.

Instead of attempting to replicate all the filters, columns, summarizations, and sorts in an Extract Step, you can simply save the existing Table Explorer settings as a new Extract Step.

Save as View

Similar to saving the current Table Explorer settings as an Extract Step above, you can also save the settings directly as a view.

This can be particularly useful when trying to construct slices of data for reporting or other downstream processes that don’t require a a data pipeline.

Manage Saved Filters

You never have to lose your filter work. You can save your Table Explorer settings as a saved filter. Saved filters also include column visibility, summarizations, columns filters, advanced filters, and sorts.

You can also let others use a saved filter by checking the Public checkbox when saving the filter.

From the Actions menu you can also choose to delete and rename saved filters.

Advanced Filters

While point-and-click column filters allow for quick application of filters to zero in on the desired results, sometimes filter conditions are complex and need more advanced specifications.

The advanced filter area provides both a pre-aggregation filter as well as a post-aggregation filter, if Summarize is enabled.

Any valid Python expression is acceptable to subset the data. Please see Expressions for more details and examples.