3 ms·

You got the right impression about our involvement, but the wrong idea about the outcome. When a professor asks you to do a semester project, you look for an a

You got the right impression about our involvement, but the wrong idea about the outcome. When a professor asks you to do a semester project, you look for an area where work within your abilities can be done. This is what we came up with. We spent a few hours searching for similar projects and found nothing, and it would have been impractical to get in touch with every maintainer of every SCM to see if they were working on a feature like revsets. We used Mercurial for two reasons: it is distributed, and it has a great hook API. There was no traffic on the mailing list that would have hinted to us that revsets were on the way. As far as I can tell, Matt Mackall himself was just as uncommunicative as we were. The only place I can find revsets mentioned prior to their introduction in Mercurial 1.6 was a mailing list post where he announced them, fully baked and implemented. Can we be expected to be more involved with the community than the primary maintainer of the project? Anyway, we never felt comfortable with sharing our work that had a nonzero probability of becoming vaporware. We only recently realized that the ideas themselves might be worth something to others, and this is our way of introducing ourselves to the community. I haven't rejoined the Mercurial mailing list yet simply because I have been busy with work and it keeps slipping my mind, but I do want to get plugged in with them.

timtadh 16y ago
(the other author here) In addition to what Steve said, this particular "feature" is really just the very beginning for us. Our ideas about what query-able version control mean go way beyond the ability to select arbitrary sets of commits based on commit meta data. We want to be able to do higher order queries. For instance: For each line of code written by Johnny what is the average life span of a line of code. Basically how lasting on average are lines of code written by a particular developer. You could combine this query to be only on a particular subgraph forest of the repository (for instance a branch, or commits this quarter). These types of queries are aggregation queries, however there are other classes of queries we would like to support as well. See some of the papers we wrote on scribd for more details. We will be writing a another post soon covering our vision for version control query.

timtadh 16y ago

(the other author here) In addition to what Steve said, this particular "feature" is really just the very beginning for us. Our ideas about what query-able version control mean go way beyond the ability to select arbitrary sets of commits based on commit meta data. We want to be able to do higher order queries. For instance: For each line of code written by Johnny what is the average life span of a line of code. Basically how lasting on average are lines of code written by a particular developer. You could combine this query to be only on a particular subgraph forest of the repository (for instance a branch, or commits this quarter). These types of queries are aggregation queries, however there are other classes of queries we would like to support as well. See some of the papers we wrote on scribd for more details. We will be writing a another post soon covering our vision for version control query.