miniblog.

Merge any sequence of patches (e.g. rebasing a branch) feels like it relies on hope. Whether or not you get conflicts, whether or not the output is syntactically valid or does the right thing: sometimes the computer does too little, and other times it does too much.

A significant part of development practice is trying to work out "should I manually change this, or should I write a complex editor macro / sed/awk script / program using an AST and refactoring library?". It's easy to make the wrong decision in both directions IME.

One of the most un-Emacsy things I've done to my configuration is binding C-x C-g to 'find all files in this repo' command. It's really useful (right next to C-x C-f!), but binding C-g (quit) feels like bad practice. Emacs doesn't judge though :)

It's such a shame there isn't structured data in commit messages. I can't query for the commit that fixes the most bugs or filter commits to only show feature additions. We need a JSON of commit message syntax so we can start building tools.

Bitbucket is dropping support for Mercurial: https://bitbucket.org/blog/sunsetting-mercurial-support-in-bitbucket They report declining usage with <1% of new users using it. Given the additional work required to support both, I understand. VCS variety is worthwhile though: perhaps hg-only platforms are easier to support.

GitHub is adding more permission levels! You can now give people the power to close issues without giving commit rights: https://github.blog/changelog/2019-05-23-triage-and-maintain-roles-beta/

It's fascinating to see the adversarial relationship between the developers of Incognito Mode and those trying to detect it. JS is a very rich environment with a ton of options. https://www.bleepingcomputer.com/news/google/google-chrome-incognito-mode-can-still-be-detected-by-these-methods/

I think it's only a matter of time before we have programming languages that hard enforce a standard formatting. For example, suppose a compiler only accepted formatted code when it's optimising.

This is a really neat alternative to find: it honours .gitignore and makes it way easier to combine filters (tricky in find). The docs argue SQL syntax is easier, which I totally agree with. I do wonder if SQL is ergonomically optimal though: is it the qwerty of query syntax? https://twitter.com/xenozoid/status/1161210113669156864

Technology seems to tend toward integration: my desktop used to have a separate smartcard, modem, network card: all now integrated hardware. https://about.gitlab.com/2018/10/16/github-launch-continuous-integration/ talks about a similar effect in GitHub: from hosting code to issues+wiki+projects+security+CI.

Excellent article on level design in games to encourage player exploration, and considering the movement motivations:

Typing out a Python program in vim, using purely speech recognition! https://youtu.be/ddFI63dgpaI The tool used in the video is Talon:

One Phabricator feature I really miss in GitHub is persistently associating commits with PRs. Phabricator commit messages have a specific syntax to show what change they're associated with. You can amend the commit without losing that metadata.

Badges to Acknowledge Open Practices: A Simple, Low-Cost, Effective Method for Increasing Transparency https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002456 'Open Data' badges on scientific papers hugely increased the amount of open data. Reminds me of READMEs with testing badges!

Foldit is an online game where users compete to fold the structures of proteins. It's remarkable to see an example of a brute compute tool (Rosetta@home) being replaced by human analysis! I generally expect to see the reverse, especially as ML grows.