Comments on Fernando Perez: Reproducible Research at the AAAS 2011 meeting in ...

@jrovegno, no abandone el esfuerzo, pues estas ide...

2011-04-19T10:36:25.405-07:00

@jrovegno, no abandone el esfuerzo, pues estas ideas se van abriendo paso poco a poco y con seguridad que van penetrando y logrando cambio.

@ondrej, thanks for the tip! I keep learning usef...

2011-04-19T10:32:15.007-07:00

@ondrej, thanks for the tip! I keep learning useful git tricks :)

@fernando: the --word-diff is new, but --color-wor...

2011-04-15T18:34:00.946-07:00

@fernando: the --word-diff is new, but --color-words is very old, I think I use it for at least a year.

Muy interesante profesor Perez. Durante el 2008 yo...

2011-03-01T13:53:40.862-08:00

Muy interesante profesor Perez.
Durante el 2008 yo le propuse a mis profesores, acá en Chile, una idea similar a la que expone, relacionada a las lecciones del Opensource para enseñar Ingeniería{1}, la verdad es que no me entendieron y me ignoraron.
Pero al menos ahora veo que no soy el único que piensa este tipo de cosas.

Saludos

Referencia
{1} Collaborative fast learning

@chuck, the sad thing about the CRU mess is that i...

2011-02-21T21:20:53.859-08:00

@chuck, the sad thing about the CRU mess is that it's par for the course in a lot of computational scientific work. And it's not an issue of malfeasance or fraud, rather that the structure of incentives and the culture of publication without explicit access to the raw data and code makes this kind of thing the almost inevitable outcome.

@dlaxalde, thanks for the --word-diff tip! It seems I need to update to 1.7.2, which isn't in Ubuntu 10.10 yet, but I'll do a local build later. It's great when you complain about a missing feature, just to find out it was implemented in the very latest release :)

[Git] doesn't work as conveniently when tra...

2011-02-21T07:42:32.206-08:00
[Git] doesn't work as conveniently when tracking for example changes in a paper (even if written in TeX), where a small change can reflow a whole paragraph, showing a diff that is much larger than the real change.

Git has a word-diff mode that is very useful for such things. Try 'git
diff --color-words' for instance. You might want to define the following
aliases in your ~/.gitconfig:
wdiff = diff --color-words
wshow = show --color-words

Very interesting. I've often wondered how well...

2011-02-20T11:16:25.414-08:00
Very interesting. I've often wondered how well validated a lot of the old fortran models were. And the mess of the CRU code was an eye opener. Reproducing results, even when done by the same people who did the original work, can be a real job given the lack of documentation and lax code versioning that is usual when results have to go out on a schedule.