At Xobni I singlehandedly wrote 30,000 lines in three months.

That fucking hurt.

But that sunk cost was my #1 motivator during the first year. I remember telling myself that I had to make this work, because if I didn't then I'd have to start all over again.

There was a point a long time ago where Xobni was very close to dying. PG said to me "Well if xyz happens then Xobni will be dead."

I was hesitant to argue against PG in those days, but I remember saying without hesitation "No, if xyz happens then I'm moving to Texas to live with my parents and eat Ramen noodles for as long as it takes to come back. I will not let this thing die." I couldn't bear the thought of starting over.

thanks for sharing this

Come to think of it, I believe the guys who wrote the most code this past summer were the Anywhere.FMs, who also got bought in the first 6 months. So while I would not want to encourage people to equate LOC with quality, it does seem to be a good sign when founders write a lot of code.

FounderLOC (FLOC?) is probably a better metric than Enterprise LOC (ELOC). Can we add this qualifier to the YCNews vocabulary?

A one liner for recursive line counts (explicitly allowing certain extensions):

    find . -regex ".*\.\(js\|py\|rb\|php\|html\)" | xargs wc -l

You can also exclude certain file patterns by throwing in a grep -v or two, eg. my command for counting JavaScript lines and excluding 3rd-party libraries is:

    find . -name "*.js" | grep -v jquery | xargs wc -l

it is a shame that find . -name "*{js,py,rb,php,html}" | xargs wc -l

I guess {} globbing is bash specific?

55,000 lines of code in the Rock Solid Arcade games so far (4 games, including one bigger unreleased in development). Started work around 3 1/2 months ago. I'd have written them in fewer lines, but I didn't have the time.

If you use Subversion, I suggest svnstat: http://sourceforge.net/projects/svnstat

We've gone from about 20k to 45k LOC in 2 months.

We're at 38,649 lines of python code according to that script excluding test code, html, css and images since January 1st, 2008.

However, this includes white space and comments. So probably 1/2 or 2/3 of that.

You don't have to guesstimate.

http://www.dwheeler.com/sloccount/

http://cloc.sourceforge.net/

Nice finds. Would have saved me writing that script and seems like it does a more complete job.

I had assumed the Zenters had written their code in Java when I read that.

What a tricky metric.

Here's a hypothetical example of what often happens to me:

1. Have a problem, need a module, copy a template: 100 LOC.

2. Add meat to the bones: 500 LOC.

3. Start testing & find gaps: 700 LOC.

4. Keep testing until it works (almost) perfectly: 1000 LOC.

5. Refactor, rename variables, tighten comments, pull common objects, strip out what was never really needed: 300 LOC.

A lot of work to get to 300 LOC, but I couldn't have done it without writing 1000 LOC in the first place. How do I measure that?

It usually takes more than three weeks to prepare a good impromptu speech. -- Mark Twain

Writing, when done well, takes time. Some journalists write to a certain length and then shorten to improve quality. A similar process occurs when programming. A scaffolding is required and the final program is smaller. Unfortunately, this process of honing is greatly undervalued and often skipped. It takes significant time and creates less impressive statistics when you've finished. Regardless, the reduced volume of code will be create a long-term saving.

  Code LOC: 1013     Test LOC: 268     Code to Test Ratio: 1:0.3

3 weeks in. But if I count all the prototypes it's probably closer to 5k lines of code. We probably need to write more code.