cuvner 0.1.1

cuv graph

This displays all the files in your project and a histogram-like graph of their coverage. Each character represents 8 lines of code, and uses a group of unicode characters (0x2580 through 0x2587) to draw a little graph. So, if those 8 lines are not covered at all, the graph will be all red; if they’re all covered, it will be all green. If 2 out of the 8 lines are covered, there will be about 25% green and the rest red.

The total size of each file can thus be easily seen (by the length of the histogram part, which wraps to subsequent lines if needed) and an idea of which parts are covered is given.

TODO:

• testing on more terminal types

• how does it look when using something besides Solarized Dark?

• useful, beyond eye-candy?

cuv lessopen

This command is intended to be used via the LESSOPEN environment variable, which lets you pre-process files that are opened with less. So, once set up (see the help via cuv lessopen --help) you can simply run less on any file in your project, and it will get syntax-highlighted and show you the line-by-line coverage with a leading green or red mini-verical bar and red background (for uncovered lines).

A header appears at the top showing the total coverage for this particular file.

TODO:

• probably the “proper” way to do this is via a Pygments plugin or extension of some sort

• option to change which Pygments style is used

• dark/light background option?

cuv spark

This shows a “spark-line” sort of thing in the console. It’s not very useful for big projects (e.g. Twisted), but gives a very quick overview of the coverage in a small amount of space. Using the same unicode characters as cuv graph, this represents each file as a single character, and its percentage coverage is graphed (so you only get granularity down to about 12.5%).

cuv hist

cuv hist produces an SVG-based “histogram” type of display, showing every line in all your files. Each little vertical bar represents a single line of course code, and is red if it was not covered or yellow if it was partially covered (only when you enable --branch coverage).

You may view this in Firefox, for example.

TODO:

• figure out how to display the text better

• options for the size or aspect ratio desired

• file size is huge; can we do better?

• is an image really the better way to go (size-wise)?

cuv pixel

(Very much in-progress, just proof-of-concept at the moment). Produces a very-tiny representation of every line of your code, organized into 80-character wide columns with each pixel (or 2x2, 1x1, or 2x1 square) representing a single character of source code, syntax-highlighted and coloured by coverage data. So, this shows the overall “shape” of your code along with coverage information.

For a small project (e.g. txtorcon) this works fairly well – in fact, ever “unreadably small” fonts can be used per-glyph and still fit on a single screen. For a larger project (e.g. Twisted, over 350k lines of code) this is a bit more problemmatic – however, with 1 pixel per glyph and 80-character width you can still get the entire project visualized on two screenfulls of a 3840x2160 display, albeit with a nearly 10MB png image…

In between those two, Crossbar.io – which weighs in at ~112k lines of code – can easily fit on a single macbook display.

Still, I’m not sure if it’s really useful but does look kind of neat. What would be useful is to have more interactivity – e.g. a single-screen “overview” that had mouse-over effects to zoom in on particular bits of code, shown in a “real” font…

cuv html

This is a rough proof-of-concept of the “interactive” ideas discussed above in the cuv pixel. What it does is render each covered file into a syntax- and coverage- highlighted PNG image and spits out an “overview” HTML page with javascript that can zoom in on the lines as you mouse-over them.

Now, this would almost certainly work way better and faster for local developers as a GTK+/Qt “native” application – but a Web-based thing was fairly fast to prototype, and has the advantage that you can publish it easily as part of your project for visitors who probably don’t want to try and install a Python GUI application…