Coding for cases you know how to handle. e.g. I want to display file contents but have fallback text until it’s available. Vs let unexpected cases crash. e.g. I must read a file to proceed. The file server connection broke. Let it crash, process restarts in a known good state, future calls may work.

Usually `Stream.unfold/2` or `Enum.reduce_while/3` works well enough for me when I find myself wanting something similar to a while loop. But plain recursive stuff can be nice too.

Really interesting comparison. Do you see an overlap between the ways the SPA world tackles CSS issues and this ecto query question? Abstractly they feel a little similar to me. In addition to CSS I also often find SPA data queries living close to calling UI component code with some utility helpers.

Sounds like you and I have different experiences. I can definitely appreciate that plenty of companies do get mired in tech debt and plenty of them in ways that could be avoided.

To be clear my point is that I do see people essentially throw away POCs all the time and I agree that type of quick iterative behavior overlaps a lot with company success. I didn’t intend to suggest anyone add anything to anything. Just reacting to the claim that POCs are always stacked on.

I find product iteration in seed through series B-ish startups is refactoring and rewriting most POC code. The thing I keep is an understanding of the business space.

The framing and design of business code is most useful early as a guidepost for new and junior hires if hiring kicks up.

I’d be surprised if that made a big difference. Every JS dev I’ve seen mix a non-JS ecosystem replaces the out-of-the-box integration no matter what it is. IMO making it easy to replace is more important to adoption than aiming for a friendly integration.

I’m open to being surprised though.

I think there was but it was deprecated (discussion linked from the old check page) hexdocs.pm/credo/Credo....

The story of Jan 6 is like this now. Trump supporters throw out several different and sometimes contradictory explanations: that it was a just cause, that it wasn’t a big deal, that the FBI did it. The point isn’t to convince but to sow doubt in the truth… of an insurrection attempt incited by lies.

Also several junior devs I’ve worked with struggled with merges conceptually when learning git and found it easier to learn with rebase or even cherry picking and applying because they felt they understood what was happening better. Which again speaks to tooling friction to me.

In workplaces I’ve mostly seen UX as an impediment to people using merges. If `--first-parent` or something similar was the default in more tools I bet there would be much less pushback from people expecting a linear history.

Nice to have some company up on this hill :)