Sort CSV

Sorting a CSV should move complete records together. A price must remain attached to its product code, and a description must remain attached to its identifier. TableWorkbench sorts rows using one or more selected columns, retains stable order for ties, and leaves stored cell text unchanged. You can choose text or exact decimal interpretation separately for each sort field. That choice affects comparison order rather than converting the output values into a different representation.

Import and confirm the file before choosing a sort. Check that the header is correctly recognized and that quoted delimiters or multiline cells have not been split into separate fields. Identifiers are read as strings, including leading zeros and long sequences of digits. Once the columns are confirmed, add a sort field and select its direction. Add more fields when the first comparison can tie and another field should determine the order within those ties.

Sort fields are evaluated in their displayed order. The first field decides whenever its two compared values differ. The second field matters only when the first is equal, and later fields resolve remaining ties. Use the up and down controls to change this priority. Sorting first by Region and then by Price groups regions before ordering prices within each region; reversing those fields gives a different global sequence and can interleave regions.

When all selected comparisons tie, the earlier relative order is preserved. This stable behavior is useful when the incoming sequence already has meaning, such as the order of records appended from suppliers. It also makes repeated runs predictable. Stability does not establish which tied record is authoritative, and sorting does not remove duplicates. If you later use Keep first or Keep last deduplication, remember that changing order can change which occurrence that separate policy retains.

Text mode compares strings using the selected sorting locale. The locale is a data-processing choice independent of the interface language. Switching the site from English to Simplified Chinese does not change the selected text ordering rule. This avoids unexpectedly reordering a dataset merely because another reviewer prefers a different language for buttons and help. Choose a locale appropriate for the text you need to arrange and inspect representative accented or non-Latin values.

Text mode is not numeric mode. A sequence of digit characters may be ordered lexically rather than by quantity, which can put 10 before 2. It is often appropriate for codes where every character is part of an identifier. Leading zeros remain meaningful, and the tool does not remove them to create a numeric interpretation automatically. Mixed-language text can have ordering subtleties, so keep the selected locale with any explanation of the result.

Numeric mode accepts decimal text using a dot as the decimal separator and no thousands separators. Optional signs and fractional parts are supported. Values that do not conform to that syntax block the operation with a source location and column, instead of being silently treated as zero. Dates, currency symbols, grouped numbers, scientific notation, and locale-specific comma decimals are not guessed into numeric values by this operation.

Comparisons use decimal arithmetic rather than converting all values through ordinary floating-point numbers. This lets neighboring long integers such as 9007199254740992 and 9007199254740993 retain their intended order. The original strings still appear in the output. A numeric comparison can treat 2 and 2.00 as equal for ordering while keeping each spelling intact. Stable tie handling then preserves their previous relative order unless a later selected field resolves the tie.

Empty values can be placed first or last. This applies independently of whether the nonempty values are ascending or descending, so the position of missing information remains an explicit choice. A cell containing the text zero is not empty. The strings NULL and N/A are also not empty, and a whitespace-only string is not automatically discarded. If those are intended missing markers, prepare the field through a visible cleaner rule before applying numeric sorting.

A failed numeric parse leaves the source and previous accepted workflow state intact. Inspect the reported value and decide whether the sort should be textual or whether a validated conversion is required. Removing a currency symbol or replacing a decimal separator can change meaning when the source conventions are mixed. The sorter deliberately avoids doing that cleanup on your behalf. Keep type interpretation and text rewriting as distinct decisions.

Take five records with codes 001 through 005 and quantity values 9007199254740993, 9007199254740992, 2, 10, and 2. Choose quantity as an ascending numeric field. The expected code order is 003, 005, 004, 002, 001. The two quantity-two records keep their previous relative order. The very large quantities remain distinct, and both the codes and quantity strings are exported exactly as stored unless you enable a separate export-protection transformation.

Add a second ascending text sort on SKU if you want the two equal quantities ordered by code under an explicit rule. In this example the visible outcome happens to be the same, but the policy is now different for another input order. Run and review the full result, then apply it to the workflow. Undo restores the prior sequence by recomputing accepted steps from the preserved source. No cells are detached from their records during either direction of that operation.