squash vs edit workflow: what "reified" them?

[nanoni17728]

I have read Steve's book on Jujutsu, spent some weeks using jj, and I want to write about my personal experience and the workflow I ended up with as I have ADHD with severe executive dysfunction.

Looking back to what Steve wrote, I now see why people would recommend using squash as their main command, or prefer working around edit. I think, but I am not sure, that the book isn't the first piece of writing that solidified squash/edit as "the" way to explain how people interact and teach how to work with jj.

Hence my question: how did we end up here? As in, why are we now talking about a "squash workflow" or "edit workflow". In a way, what I do is slightly different to these, since I cannot "just" do these little ceremonies cleanly. I have to manage my mess somehow because that's how it always end up and I would like to label that properly so that it can be later referenced easily.

In any case, I will spend some time sharing my draft for proof reading to make sure I am not yapping misinformation.

@steveklabnik I hope you don't mind me pinging you. Thank you for writing the book :)

[jennings]

I think this is just a natural consequence of jj automatically committing all changes.

When you want to amend a revision, you can run jj edit REV. Any changes you make to the working copy are immediately (on next jj invocation) amended into the working copy commit and descendants are rebased on top. This is "the edit workflow" because it starts by running jj edit on the revision you intend to amend.

But, for many reasons, you may want to make changes without immediately amending them into the revision you ultimately want to update. So, you instead start with jj new REV, giving you a clean working copy where you can make and discard changes. When you're happy with them, you run jj squash to move them into the revision you originally wanted to amend. This is the "squash workflow" because it ends by running jj squash.

[joyously]

I haven't read the book you are referring to (although I don't know since there's no link), but I think it should be made clear that you can also use jj quite nicely without ever using either squash or edit.
However, this usage would mostly be in a solo project. When collaborating with others, the review process dictates that a submission be changed one way or another.

[nanoni17728]

I totally forgot to add the full context, my bad! Here is the link.

Indeed, jj is flexible enough to allow people to work the way they want to. This is what prompted me to open this discussion. It didn’t feel quite right that squash or edit were presented as the only central commands. I had to dig deeper.

Thank you for reminding me that by working with others, you will have to amend the changes you want to submit, meaning that squash, edit, and some others, will have to be part of the commands you frequently use. I mostly am a solo dev for now, lacking this perspective.

[steveklabnik]

I don't mind you pinging me but I'm a bit bad with notifications sometimes, sorry bout that!

As in, why are we now talking about a "squash workflow" or "edit workflow".

When I talk to people about their workflows, these are the two most common that come up. jj has a lot of flexibility, but these two just seem to be pretty common.

[nanoni17728]

Well, in the end I’m the one who is replying late, so no worries! Thank you for your input. It sounds trivial after the fact, but I was too caught up tunnel-visioning on a prescriptive explanation that I forgot the descriptive angle.

[gaojunran]

The two workflows talk about one thing: how to commit. In my way, I use jj commit --interactive --message "message", which is quicker and meets my requirements that I want to give a commit message after my work.

I'm working on hj, which is a toolbox for jj, hope to gain a star if you like! (docs are wip)

[nanoni17728]

Starred! ✨

[nanoni17728]

I wish I could spend some time to do a summary and add closing thoughts, but right now my brain is just not working for a task like that. In the meantime, I think Jenning’s answer is the most detailed, which I’ll will mark as being the "correct" answer for the time being, even though Steve’s and Gaojunran’s point in the same direction.

Thank you all for your time :)