Open-Gridview - the FOSS Out-Gridview

3 minute read

The Story

I’ve been a Microsoft SysAdmin for a long time before switching for Linux. During which, I scripted a lot in PowerShell.
PowerShell has several “output” functions like Out-File (which passes the input to a file), Out-Null etc.
One of these functions was Out-Gridview which tabularizes its input and displays it in an interactive table:
You can see how easy it is to tabulate your objects.

The advantage might be hard to explain to hardcore *nix CLI users - they might scoff and say that you can filter using grep, sort using sort and extract columns using awk, and that all output is beautiful in its own way.
However, since switching to Linux I’ve had numerous incidents of either trying to format my output so that it is both readable and parsable, or piping output to some filter and regret filtering it. How I missed being able to pass the output to Out-Gridview and being able to manipulate it via GUI.

The Solution

I’ve decided that Out-Gridview is a basic enough problem so I can tackle it, and also writing UI is fun (as if), so I implemented it via GTK and Python.
The result is the package opengridview.


For regular usage, use the pip package:

sudo pip install opengridview

For modification and hacking, set up a venv in the git repo:

git clone .
virtualenv . --system-site-packages
. bin/activate
pip install -e .

Basic usage

To use it, pipe your output into ogv (or open-gridview), optionally specifying how to parse each row (the default is “autosplit”) and how the columns are named (usually the first row is taken as a header row).
ogv can accept regular files, but it’s more interesting to pipe streams into it.
Multiple files can be passed (causing multiple windows to appear) by using bash’s process substitution:

ogv <(ls -l /home/bob | tail -n+2) <(ls -l /var | tail -n+2) --headers permissions,links,user,group,size,mod-month,mod-day,mod-time,name



Emulates perl’s autosplit, using re.split and a delimiter (defaults to \s+)


Uses the csv module to prase every line. Configurable delimeter (defaults to ,)


Treats each line as a separate JSON object

Neat features

  • The autosplit parser uses re.split to create the columns. This method honors capturing groups inside the separator, meaning you can use the separator as a Regex parser, and discard the “split” columns using empty column headers, like this:

      pip list | ogv -s '^([^ ]*) \(([^ ]*)\)$' --headers ,name,version

    (see below for screenshot of the result)

  • I implemented Ctrl+c (inside the GUI) to copy selected rows as CSV text, in order to ease copying data into spreadsheet applications (a common action for me when handling tabulated data)

Example use cases

Filesystem stuff

ls -l | tail -n+2 | ogv --headers permissions,links,user,group,size,mod-month,mod-day,mod-time,name &
du -h | ogv --separator '\t' --headers size,path &

Show all packages installed

Preset the filter to “python”

aptitude search ~i --disable-columns -F '%p,%v,%t' | ogv -p csv --headers name,version,source --filter python

Pip package list

Custom headers, column extraction using capturing groups

pip list | ogv -s '^([^ ]*) \(([^ ]*)\)$' --headers ,name,version

  echo platform name version
  pip list | perl -ne 'print "python $_"';
  gem list | grep '(' | perl -ne 'print "ruby $_"'
) | ogv

Interesting bits

  • I chose to implement Open-Gridview on GTK because I’m running Gnome
  • I chose Python (as opposed to Ruby / Perl) because Python comes preinstalled on most major Linux distros
  • Since I chose PyGObject as my GTK library, I’m annoyingly missing some features, like:
    • Writing custom data to the clipboard
    • Directly connecting functions to key shortcuts (I’m currently using menu items as a middleman)
    • The Gtk.Application class forces “single-instance” applications on you (think firefox). Although there was some discussion about this decision, it seems mandated. I had to work around this by managing multiple Gtk.Window instances by myself
  • Some Python utilities (e.g. csv.reader) use internal buffering when reading streams. This means that using them directly caused delay when parsing a live stream (and not a file). I’m currently reading one line at a time, then feeding it to a parser. This is also why all current parsers are line-based
  • I found no easy/elegant way to interrupt streams while they were waiting for input. Eventually I found about “daemon threads”, which are threads that don’t justify keeping the process alive (if only daemon threads are left, the process is closed)


  • Add item selection functionality (like out-gridview’s OutputMode)
  • Support advanced filtering (e.g. regex, specific columns)
  • Support hiding and re-ordering columns
  • Add a progress bar if the stream isn’t closed (EOFed)

Please let me know if you find this useful.