-
-
Notifications
You must be signed in to change notification settings - Fork 43
About commiting HTML files #159
Comments
On Thu, Feb 05, 2015 at 01:19:17PM -0800, Raniere Silva wrote:
This is information for contributors, not all users, so I'd recommend |
I think `CONTRIBUTING.md` is the right place.
|
On 05/02/15 22:31, Greg Wilson wrote:
I have been thinking about this, since I sat down last weekend to work I think it has to be up front straight away somehow, even if that just karin |
On Fri, Feb 06, 2015 at 08:19:30AM -0800, karinlag wrote:
GitHub posts a nice, highlighted link to CONTRIBUTING.md at the top |
Yup - but if people aren't seeing it, they aren't seeing it, so we have
G |
For the record, @ChristinaLK has a version of what's needed in the lessons at swcarpentry/shell-novice#68. Once that's settled down I guess it should be copied here and and made into a non-specific template (with a note somewhere in the instructions for creating a new lesson saying that the file will need updating). Where should instructions for contributing to this repository go? |
On 06/02/15 18:50, Greg Wilson wrote:
This is after you have opened an issue or a pull request. You have to
If you see right away that all you have to do is to fix the md, it K |
I'm convinced - please submit a PR to add something prominent to README.md.
|
On Sun, Feb 08, 2015 at 04:17:57AM -0800, karinlag wrote:
Then maybe we need to be more welcoming, since “there's a typo on page We welcome contributions of any kind. For more information see |
On 08/02/15 17:32, W. Trevor King wrote:
Hm, I might not have dug enough, but I could not find a template README K |
On Mon, Feb 09, 2015 at 11:28:34AM -0800, karinlag wrote:
I expect it should go somewhere like this 1. |
I'm closing this since now we have |
This model of only committing the source files for a PR works when someone is proposing a small change. But what about when someone creates many changes? It has been my experience that most instructors (non-trainees that is) make changes when they are preparing for a workshop. What is the protocol then? Should they always do separate commits whenever they make a change (one for the source and one for the html)? And then when they submit a PR, they'll have to somehow cherrypick only the source commits? Or if they think ahead, they could make all the source commits in a separate branch and then merge them into gh-pages for their workshop (but that would take a lot of foresight on their part since we don't have any instructions that they should do that). As a concrete example, see this PR to r-novice-inflammation. Because of all the recent development, there are tons of merge conflicts. But we can't separate out the Rmd from the html and md because the changes were committed at the same time. |
On Mon, Feb 16, 2015 at 09:36:53AM -0800, John Blischak wrote:
I think so. In other projects, folks often have separate master and
Yes, and “somehow” is probably just going to be “if I'd realized I $ git checkout -b my-feature origin/gh-pages # or whatever the upstream lesson is which will adjust you cherry-pick to restore the version of the HTML
This is easier than reconstructing those feature branches after the $ git add -p all the unsharable work, then $ git stash the shareable work to a new feature branch based on the upstream $ git checkout gh-pages to pull those source-only commits back into your per-workshop lesson That's not a very complicated workflow, but it requires a good deal $ git log --graph --topo-order --oneline --decorate --all (I keep most of that in an alias 1) should be sufficient to keep |
Thanks for the advice, @wking. I agree these would be good ways to develop the material if everyone had a decent understanding of Git. But the reality is that most of our instructors do not, and we have no requirement of knowledge of branching as part of instructor training. I want to write better documentation to make my life easier when contributors send PRs to the repo I maintain, but I'm not going to be able to write documentation that can also teach them how to use Git in the first place. |
On 2015-02-16 11:28 PM, John Blischak wrote:
|
But Greg, my comment was in response to your answer to the thread that started this Issue. You want development in only one branch, but then you want PRs to only include changes to the source code. Then any changes someone makes to improve the lesson for use in their workshop will be lost unless we use one of the strategies that @wking described above (because they will also have to generate the html for their workshop). There is a disconnect here. I want to improve the documentation for contributing to the R lesson, but I can't do that when even I am unclear of what the official process is (before this thread I was having contributors include the generated content in their PR). |
On Tue, Feb 17, 2015 at 07:18:20AM -0800, John Blischak wrote:
For folks who aren't ready to use a feature-branch workflow locally,
This still works, it just has a higher risk of conflicts in the |
@wking , I just don't see how this approach scales. SWC is growing larger by the day. We're increasing both the number of distinct lesson repos and instructors. For SWC to grow into a large, distributed, open-source project, we need reliable documentation that any new collaborator can read and then start contributing. Determining how to merge PRs on a case-by-case basis is too time-consuming. P.S. Perhaps this problem is only particularly acute for the R lessons. Because we write in R Markdown and thus re-run the code when we build a file, there tend to be lots of changes to the html (e.g. because different versions of R have slightly different wordings of error messages). This of course creates many unnecessary merge conflicts. |
On Thu, Feb 19, 2015 at 07:07:21AM -0800, John Blischak wrote:
The approach that scales is getting contributors to create source-only
That's the basic workflow that gets maintainers something they can 3a. ‘git checkout -b my-feature’ to collect your source changes in Once they get comfortable with 3a, they can improve to 3b: 3b. ‘git checkout -b my-feature upstream/gh-pages’ to root your source
I don't think my suggestions one and two are beyond anyone's
No matter how good your docs are, some folks will fall through the |
I tend to agree with @wking. For one imho one of our goals should be to make people more comfortable with git to a level they can follow the process outlined above. I also agree that it's not too hard to rebase someone else's pr to make it conform to these rules. I have done similar git surgery many times and I assume so have many of the more senior SWC contributors. |
The discussion of this issue happened at our mailing list, see http://lists.software-carpentry.org/pipermail/discuss_lists.software-carpentry.org/2015-February/002573.html.
@jiffyclub asked
@gvwilson Where can we write the answer for this question?
README.md
?The text was updated successfully, but these errors were encountered: