I’m currently in a project team working on an application that stores much of its data in CouchDB. One of the lovely things about Couch is its RESTful API. It’s all simple HTTP and JSON, easy to understand and easy to program to.

One aspect where this interface isn’t so readily accessible is in shell scripting. There’s curl to handle all the HTTP stuff we could ever need, but to transform a JSON structure or extract information from it proved less straightforward. We can cover simple cases with grep and awk, but JSON is complex enough that we (or, well, I) wouldn’t want to. If the documents were XML, we could have used xpath and xslt to do our heavy lifting. There is to my knowledge no equivalent to xmlstarlet for JSON to reliably handle these chores.

We solved our shell scripting problem and the solution is dead obvious,  Read more