Filter CSV

Filtering CSV means selecting complete records according to rules, while keeping a separate account of the records that do not qualify. TableWorkbench supports text conditions on chosen columns and lets you combine them with Match all or Match any. It does not turn every value that looks numeric into a number, and it does not run arbitrary expressions. The goal is a selection you can explain: which fields were checked, which tests were applied, and which records were retained or excluded.

Choose a file or paste a table, then confirm parsing before creating conditions. Check the column names, record count, and representative cell values. If a row is shorter than the declared structure, you must explicitly permit padding with missing cells. Extra fields and malformed quoting block parsing. These decisions happen before filtering because a misplaced delimiter could otherwise make a condition evaluate the wrong field while still producing a plausible-looking subset.

Each condition identifies one column and a text test. Equals requires the complete stored text to match the supplied value. Contains checks whether the value occurs anywhere in the cell. Starts with and Ends with constrain the position of the text. Is empty matches empty or missing cells, while Is not empty selects populated cells. Empty tests do not need a separate comparison value. Choose the operator that expresses the actual selection, rather than approximating it with a broader substring test.

For example, Contains North also matches North-East, whereas Equals North does not. Starts with AB can select a family of product codes while leaving codes with AB only in the middle excluded. The tests use the original strings and are case-sensitive in this version. If the source needs standardized case or whitespace, add a deliberate cleaning step first. That keeps text changes separate from the later inclusion decision and makes both stages reversible.

Match all retains a record only when every configured condition succeeds. Use it for a question such as Region equals North and Status equals Ready. Match any retains a record when at least one condition succeeds. It can select records from either North or South when both equality conditions refer to Region. The visible order of conditions helps you review them, but these simple logical combinations do not assign a higher priority to the first rule.

This version does not provide nested groups such as A and either B or C in one expression. You can sometimes use separate applied filtering steps to express a narrower sequence, but check the resulting logic carefully: applying one filter after another behaves like further restricting the previous result. Keep the excluded report from each stage if you need a complete explanation of where records left the workflow. Do not use a sequence as a substitute for an unverified complex expression.

An empty cell is not the same thing as the text NULL, null, N/A, false, or 0. Is empty matches the empty and missing states represented by the table; those literal strings remain content. A cell containing only spaces also remains content until a cleaning step changes it. This matters when a source uses several conventions for missing data. Decide which conventions should count as missing and transform them explicitly before filtering, if necessary.

Numeric and date comparisons are outside this version's text-filter semantics. The strings 10 and 2 are not silently interpreted as quantities for a greater-than test, and an ambiguous date is not guessed from the interface language. Use Equals when you truly want the literal text 10. If you need range filtering, define the field's type and parsing rules in a suitable external process or wait for an explicitly supported typed operation. The current interface does not claim that capability.

Imagine four records with SKU, Region, and Status. SKU 001 is North and Ready; 002 is North and Hold; 003 is South and Ready; 004 has an empty region and Ready status. Add Region equals North and Status equals Ready, then choose Match all. Only 001 qualifies. The retained count is one and the excluded count is three. All columns of the retained record stay together, including fields that were not part of a condition.

Change to Match any and run again. Records 001, 002, 003, and 004 now qualify because every record either has Region North or Status Ready. This is a valid outcome even though it may be broader than intended. The example shows why changing all to any is a substantive rule change rather than a display preference. Inspect both the summary and excluded table; an empty excluded report can be meaningful evidence that the conditions cover the entire input.

The search box above the data grid only changes which rows you see in that current view. It does not add a workflow condition, change the retained count, or narrow a download. This separation lets you inspect a particular SKU without accidentally exporting only that one row. A real filter is created in the rule panel, run across all imported records, and optionally applied as a workflow step. Its included and excluded outputs are explicit.

After reviewing the result, apply it to continue with the retained records. Undo returns to the previous table, and redo can restore the filter. If you change a condition, the old result is marked stale until another run completes. Export asks for a scope independent of grid search: choose the retained result, the excluded report, or a package of reports. This makes it possible to provide an exception list without losing the accepted dataset.