CSV Splitter for real datasets

Rows per file gives every output roughly the same row count. This is best when downstream tools care about batch size, not category or meaning.

Overlap rows repeats a few rows at the boundary between files. Use it only when the next batch needs a little context from the previous one.

Keep grouped rows together stops the tool from breaking the same account, company, or order across several files. It matters when rows belonging to one entity should stay together.

Max size per file is for upload limits. It aims to stay under a byte target, but row lengths can vary, so the last file or some files may still feel less even than row-based chunks.

Split by column(s) creates one output per distinct value or value combination. One column gives simpler outputs; two or more columns create more granular files.

Rare-value handling helps when one-off values would create too many tiny files. You can keep only the biggest groups and merge the rest into an “other” bucket.

Also split large groups by file size is useful when one category is much bigger than the rest. It keeps the grouping logic, but prevents a single giant file from dominating the export.

First match only means a row stops at the first rule it matches. Turn it off only when you intentionally want the same row copied into several output files.

Keep unmatched rows in a separate file makes sure rows that satisfy none of your rules are still preserved in their own export.

Unmatched rows file name is simply the name of that catch-all file. It helps when you want a clear output like needs-review.csv or no-rule-match.csv.

Date format is there for ambiguous date columns. If the detector guesses wrong, force the expected format before grouping by month, year, or week.

Invalid dates lets you decide whether bad values should be skipped or moved into a separate file. That is safer than silently losing rows.

Filename pattern controls final output names. Use variables like {index} and {label} when your batches need predictable numbering.

Columns to keep in the output removes unnecessary columns before the split. That keeps exports lighter and cleaner, especially when the original CSV has helper columns you do not want downstream.

The parser keeps quoted CSV values intact, so addresses, notes, or long text fields do not get broken into fake rows during the split.

Duplicate headers are normalized so comparisons still work. Empty columns can be identified, and projection lets you remove columns before generating outputs.

For large row-based splits, the tool can keep preview lightweight and only preload the first output file. Secondary previews and downloads are materialized on demand.

Rows with invalid dates can go into a separate file, rows that match no rule can be kept in their own export, and grouped splits can warn you when one segment is much larger than the rest.