distributed spreadsheets with intelligent merges
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
scripts update mingw build for current gnumeric Jun 27, 2014
ChangeLog bump version to 0.6.0 Feb 13, 2012


The COOPY toolbox

Build Status

Diffing, patching, merging, and revision-control for spreadsheets and databases. Focused on keeping data in sync across different technologies (e.g. a MySQL table and an Excel spreedsheet).

The main programs

  • ssdiff - generate diffs for spreadsheets and databases.
  • sspatch - apply patches to spreadsheets and databases.
  • ssmerge - merge tables with a common ancestor.
  • ssformat - convert tables from one format to another.
  • ssfossil - the fossil DVCS, modified to use tabular diffs rather than line-based diffs. You can also work with git. If you use github, you may want to check out CSVHub, which uses a simplified version of ssdiff called daff to show pretty data diffs on github.
  • coopy - a graphical interface to ssfossil.

Supported data formats

  • CSV (comma separated values)
  • SSV (semicolon separated values)
  • TSV (tab separated values)
  • Excel formats (via gnumeric's libspreadsheet)
  • Other spreadsheet formats (via gnumeric's libspreadsheet)
  • Sqlite
  • PostgreSQL
  • MySQL
  • Microsoft Access format (via mdbtools - READ ONLY, or via jackcess for read/write)
  • A JSON representation of tables.
  • A custom "CSVS" format that is a minimal extension of CSV to handle multiple sheets in a single file, allow for unambiguous header rows, and have a clear representation of NULL.

Supported diff formats

Example uses

  • Enumerating differences between any pairwise combination of CSV files, database tables, or spreadsheets.
  • Applying changes to a database or spreadsheet, without losing meta-data (formatting of spreadsheet, indexing/type information for database). Particularly useful for applying changes in an exports CSV file back to the original source.
  • Editing a MySQL/Sqlite database in gnumeric/openoffice/Excel/...
  • Distributed editing of a spreadsheet/database using a DVCS. Benefits: revision history, offline editing in tool of choice, self-hosting possible.


  • By default, when comparing tables, no initial assumption is made about schema similarity. Column names are not required to exist, or to be preserved between tables. The number and order of columns may also differ.
  • If schema changes are not expected, COOPY can be directed to use certain columns as a trusted identity for rows (a key).
  • Respects row order for table representations for which row order is meaningful (spreadsheets, csv).
  • By default, COOPY assumes your data is very messy. If it is clean, you can get much faster results by tweaking some options.


The core of the COOPY toolbox is a 3-way comparision between an ancestor and two descendents. First, rows are compared using bags of substrings drawn from across all columns. Once corresponding rows are known, columns are compared, again using bags of substrings. Row and column assignments are optimized and ordered using a Viterbi lattice. Once the pairwise relationships between each descendent and its ancestor are known, differences are computed, and a good merged ordering is determined (again using the Viterbi algorithm).

Installing on OSX

  • Use homebrew.
  • Do brew tap paulfitz/data to get a formula for coopy.
  • Install XQuartz from http://xquartz.macosforge.org
  • Then brew install coopy should work fine.

Installing on Windows

Installing on Linux

  • Sorry, this is where I develop myself, but I don't have an installer. Building is easy though!


  • For a stripped-down js/py/rb/php version see http://paulfitz.github.io/daff/
  • See BUILD.md for information on building the programs.
    • Summary: CMake
  • See SERVE.txt for server-side information.
    • Summary: fossil
  • See COPYING.txt for copyright and license information.
    • Summary: GPL. Relicensing of library core planned for version 1.0.


COOPY targets a stable, fully-documented release at version 1.0. At the time of writing, the version number is just beyond 0.5. It is about half way there.

Apparently COOPY is the closest thing right now to git for data:

But if you deal with big data sets and don't care so much about diffs and patches and whatnot, you may want to look at dat: