Find and replace in CSV

Find and replace in CSV should operate on parsed cell values, not on the raw file as an undifferentiated string. A comma can be a delimiter or part of a quoted description, and a newline can separate records or live inside one cell. TableWorkbench parses the table first, applies literal text replacement only to the columns you choose, and serializes the result correctly when exporting. This keeps structural characters separate from the content you intend to edit.

Choose a UTF-8 CSV, TSV, or delimited text file, or paste cells from a spreadsheet. Confirm the delimiter, quoting, header, and any skipped introductory records. If the parser identifies malformed quotes or extra fields, resolve the input rather than proceeding with an uncertain structure. Once the file is confirmed, select the columns that should be searched. No column is selected for editing merely because it contains a value that resembles your search text.

Entire cell requires the cell's complete text to match your find text under the selected case rule. It is useful for replacing a known category code or a specific missing marker without touching longer descriptions that contain the same characters. Substring mode replaces occurrences within a cell, which fits changing a repeated label or correcting a phrase in selected description fields. The two modes can produce very different scopes, so inspect them with representative data.

The find text must be nonempty. Searching for an empty string would match between characters and create a surprising bulk edit, so the tool blocks it. The replacement may be empty, which means the matching content is removed. In whole-cell mode that clears the cell. In substring mode it removes only the matched portions and leaves other text. The interface states this consequence so a blank replacement is an intentional instruction rather than an omitted field.

Case-sensitive matching is the default. A find value of blue will not match Blue unless you turn that distinction off. Case-insensitive literal matching is useful when different source systems capitalize a known label differently. It is not fuzzy matching, translation, accent removal, or spelling correction. Do not assume that two visually similar words will match under every language's conventions. Test the exact characters present in your file and inspect the reported changes.

The first version does not accept regular expressions or user scripts. A dot, asterisk, bracket, or dollar sign in the find field is literal text, not programming syntax. Replacement text is also inserted literally, including quotation marks or delimiters. This keeps a nontechnical task bounded and avoids executing expressions from imported content. Formulas and script-looking strings in cells are never evaluated by the workbench.

Suppose a record has SKU 001, a Note cell containing A, a comma, a lowercase a, a newline, and b, plus a separate Keep cell containing A. Select Note only. Search for a, replace it with a quoted Q, choose substring matching, and turn off case sensitivity. There are two matching occurrences inside the Note cell. The Keep column remains A because it was outside the selected scope, and SKU remains the text 001.

The replacement introduces quote characters into a cell that already contains a comma and newline. The internal result stores the intended text, and export adds the necessary CSV quoting and doubled quote escapes. Reading the generated CSV should recover the same cell value. A raw search-and-replace across the original file could instead alter delimiters or quoting syntax, which is why the parser and serializer surround the text-editing operation.

Matched occurrences counts how many find matches were identified. Affected cells counts cell changes, and affected records identifies rows whose stored values changed. One cell can contain several occurrences, while several changed cells can belong to one record. A replacement that inserts the same text it found can have matches without a meaningful value change. Interpret these statistics according to their labels instead of expecting all the numbers to be equal.

The change report includes the column, prior value, resulting value, and original source location. Use it to inspect the exact edit before applying the result to the workflow. The original view remains available, and a search in the preview only helps locate records; it does not change the replacement scope. If you change the find text, replacement, selected columns, or matching rule, the prior result becomes stale and must be recomputed before export.

Apply the reviewed result when the next tool should use the replacement. You can then clean another field, deduplicate by a corrected code, map columns, or split the table without downloading it between steps. Undo returns to the preceding workflow table, and redo restores the accepted replacement when available. Reordering or disabling workflow steps can change later results, so the workbench recomputes from preserved inputs rather than treating an old preview as current.

Save rules can help repeat a known replacement on a table with the same source structure. The downloaded configuration does not contain the complete dataset, but the find and replacement values can still reveal sensitive information. Check a rule file before sharing it. A loaded rule should be reviewed against the actual new data even when its headers match, because a familiar column name does not guarantee that the same replacement is appropriate for every source.

Choose the full result or the change report explicitly in the export dialog. CSV and TSV are supported, with header inclusion and UTF-8 BOM controls. Spreadsheet protection can add an apostrophe to values beginning like formulas, including some legitimate signed text. Its count is shown before download. Raw export preserves the stored strings but requires acknowledging that another application may interpret them as formulas. These export choices are separate from the replacement you reviewed.

The generated text is parsed again to verify its intended records and cells before a new file is offered. A failed verification produces an error rather than a partial success. This check supports structural correctness; it does not establish that your replacement phrase is factually correct or that a receiving system will accept the file. Keep the source and the change report when the edit needs to be explained later.