Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List More from tonsky.me
Каким должен быть редактор кода в 2024 году? Почему Vim морально устарел, а IDEA, кажется, сдает позиции? Популярность Zed, минимализм SublimeText, гибкость Emacs и многое другое в новом выпуске.
Squint is a light-weight dialect of ClojureScript with a compiler and standard library. “The idea is that when you squint, it still looks like CLJS”.
Clojure macros have two modes: avoid them at all costs/do very basic stuff, or go absolutely crazy. Here’s the problem: I’m working on Humble UI’s component library, and I wanted to document it. While at it, I figured it could serve as an integration test as well—since I showcase every possible option, why not test it at the same time? This is what I came up with: I write component code, and in the application, I show a table with the running code on the left and the source on the right: It was important that code that I show is exactly the same code that I run (otherwise it wouldn’t be a very good test). Like a quine: hey program! Show us your source code! Simple with Clojure macros, right? Indeed: (defmacro table [& examples] (list 'ui/grid {:cols 2} (for [[_ code] (partition 2 examples)] (list 'list code (pr-str code))))) This macro accepts code AST and emits a pair of AST (basically a no-op) back and a string that we serialize that AST to. This is what I consider to be a “normal” macro usage. Nothing fancy, just another day at the office. Unfortunately, this approach reformats code: while in the macro, all we have is an already parsed AST (data structures only, no whitespaces) and we have to pretty-print it from scratch, adding indents and newlines. I tried a couple of existing formatters (clojure.pprint, zprint, cljfmt) but wasn’t happy with any of them. The problem is tricky—sometimes a vector is just a vector, but sometimes it’s a UI component and shows the structure of the UI. And then I realized that I was thinking inside the box all the time. We already have the perfect formatting—it’s in the source file! So what if... No, no, it’s too brittle. We shouldn’t even think about it... But what if... What if our macro read the source file? Like, actually went to the file system, opened a file, and read its content? We already have the file name conveniently stored in *file*, and luckily Clojure keeps sources around. So this is what I ended up with: (defn slurp-source [file key] (let [content (slurp (io/resource file)) key-str (pr-str key) idx (str/index-of content key) content-tail (subs content (+ idx (count key-str))) reader (clojure.lang.LineNumberingPushbackReader. (java.io.StringReader. content-tail)) indent (re-find #"\s+" content-tail) [_ form-str] (read+string reader)] (->> form-str str/split-lines (map #(if (str/starts-with? % indent) (subs % (count indent)) %))))) Go to a file. Find the string we are interested in. Read the first form after it as a string. Remove common indentation. Render. As a string. Voilà! I know it’s bad. I know you shouldn’t do it. I know. I know. But still. Clojure is the most fun I have ever had with any language. It lets you play with code like never before. Do the craziest, stupidest things. Read the source file of the code you are evaluating? Fetch code from the internet and splice it into the currently running program? In any other language, this would’ve been a project. You’d need a parser, a build step... Here—just ten lines of code, on vanilla language, no tooling or setup required. Sometimes, a crazy thing is exactly what you need.
So I was at the Local-First Conf the other day, listening to Martin Kleppmann, and this slide caught my attention: Specifically, this part: But first, some context. What is local-first? For the long version, go to Ink & Switch, who coined the term. Or listen for Peter van Hardenberg explaining it on LocalFirst.fm. Here’s my short version: It’s software. That prefers keeping your data local. But it still goes to the internet occasionally to sync with other users, fetch data, back up, etc. If it doesn’t go to the internet at all, it’s just local software. If it doesn’t work offline with data it already has, then it’s just normal cloud software. You all know the type — sorry, Dave, I can’t play the song I just downloaded because your internet disappeared for one second... But somewhere in the middle — local-first. We love it because it’s good for the end user, you and me, not for the corporations that produce it. What’s the problem with local-first? The goal of local-first software is to get control back into the hands of the user, right? You own the data (literally, it’s on your device), yada-yada-yada. That part works great. However, local-first software still has this online component. For example, personal local-first software still needs to sync between your own devices. And syncing doesn’t work without a server... So here we have a problem: somebody writes local-first software. Everybody who bought it can use it until the heat death of the universe. They own it. But if the company goes out of business, syncing will stop working. And companies go out of business all the time. What do we do? Cue Dropbox The solution is to use something widely available that will probably outlive our company. We need something popular, accessible to everyone, has multiple implementations, and can serve as a sync server. And what’s the most common end-user application of cloud sync? Dropbox! Well, not necessarily Dropbox, but any cloud-based file-syncing solution. iCloud Drive, OneDrive, Google Drive, Syncthing, etc. It’s perfect — many people already have it. There are multiple implementations, so if Microsoft or Apple go out of business, people can always switch to alternatives. File syncing is a commodity. But file syncing is a “dumb” protocol. You can’t “hook” into sync events, or update notifications, or conflict resolution. There isn’t much API; you just save files and they get synced. In case of conflict, best case, you get two files. Worst — you get only one :) This simplicity has an upside and a downside. The upside is: if you can work with that, it would work everywhere. That’s the interoperability part from Martin’s talk. The downside is: you can’t do much with it, and it probably won’t be optimal. But will it be enough? Version 1: Super-naive Let’s just save our state in a file and let Dropbox sync it (in my case, I’m using Syncthing, but it’s the same idea. From now on, I’ll use “Dropbox” as a common noun). Simple: But what happens if you change the state on two machines? Well, you get a conflict file: Normally, it would’ve been a problem. But it’s not if you are using CRDT! CRDT is a collection of data types that all share a very nice property: they can always be merged. It’s not always the perfect merge, and not everything can be made into a CRDT, but IF you can put your data into a CRDT, you can be sure: all merges will go without conflicts. With CRDT, we can solve conflicts by opening both files, merging states, and saving back to state.xml. Simple! Even in this form, Dropbox as a common sync layer works! There are some downsides, though: conflicting file names are different between providers, some providers might not handle conflicts at all, it needs state-based CRDT. Version 2: A file per client The only way to avoid conflicts is to always edit locally. So let’s give each client its own file! Now we just watch when files from other clients get changed and merge them with our own. And because each file is only edited on one machine, Dropbox will not report any conflicts. Any conflicts inside the data will be resolved by us via CRDT magic. Version 3: Operations-based What if your CRDT is operation-based? Meaning, it’s easier to send operations around, not the whole state? You can always write operations into a separate append-only file. Again, each client only writes to its own, so no conflicts on the Dropbox level: Now, the operations log can grow quite long, and we can’t count on Dropbox to reliably and efficiently sync only parts of the file that were updated. In that case, we split operations into chunks. Less work for Dropbox to sync and less for us to catch up: You can, of course, save the position in the file to only apply operations you haven’t seen. Basic stuff. Theoretically, you should be able to do operational transformations this way, too. Demo A very simple proof-of-concept demo is at github.com/tonsky/crdt-filesync. Here’s a video of it in action: Under the hood, it uses Automerge for merging text edits. So it’s a proper CRDT, not just two files merging text diffs. Conclusion If you set out to build a local-first application that users have complete control and ownership over, you need something to solve data sync. Dropbox and other file-sync services, while very basic, offer enough to implement it in a simple but working way. Sure, it won’t be as real-time as a custom solution, but it’s still better for casual syncs. Think Apple Photos: only your own photos, not real-time, but you know they will be everywhere by the end of the day. And that’s good enough! Imagine if Obsidian Sync was just “put your files in the folder” and it would give you conflict-free sync? For free? Forever? Just bring your own cloud? I’d say it sounds pretty good.
More in programming
I was chatting with a friend recently, and she mentioned an annoyance when reading fanfiction on her iPad. She downloads fic from AO3 as EPUB files, and reads it in the Kindle app – but the files don’t have a cover image, and so the preview thumbnails aren’t very readable: She’s downloaded several hundred stories, and these thumbnails make it difficult to find things in the app’s “collections” view. This felt like a solvable problem. There are tools to add cover images to EPUB files, if you already have the image. The EPUB file embeds some key metadata, like the title and author. What if you had a tool that could extract that metadata, auto-generate an image, and use it as the cover? So I built that. It’s a small site where you upload EPUB files you’ve downloaded from AO3, the site generates a cover image based on the metadata, and it gives you an updated EPUB to download. The new covers show the title and author in large text on a coloured background, so they’re much easier to browse in the Kindle app: If you’d find this helpful, you can use it at alexwlchan.net/my-tools/add-cover-to-ao3-epubs/ Otherwise, I’m going to explain how it works, and what I learnt from building it. There are three steps to this process: Open the existing EPUB to get the title and author Generate an image based on that metadata Modify the EPUB to insert the new cover image Let’s go through them in turn. Open the existing EPUB I’ve not worked with EPUB before, and I don’t know much about it. My first instinct was to look for Python EPUB libraries on PyPI, but there was nothing appealing. The results were either very specific tools (convert EPUB to/from format X) or very unmaintained (the top result was last updated in April 2014). I decied to try writing my own code to manipulate EPUBs, rather than using somebody else’s library. I had a vague memory that EPUB files are zips, so I changed the extension from .epub to .zip and tried unzipping one – and it turns out that yes, it is a zip file, and the internal structure is fairly simple. I found a file called content.opf which contains metadata as XML, including the title and author I’m looking for: <?xml version='1.0' encoding='utf-8'?> <package xmlns="http://www.idpf.org/2007/opf" version="2.0" unique-identifier="uuid_id"> <metadata xmlns:opf="http://www.idpf.org/2007/opf" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:dcterms="http://purl.org/dc/terms/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:calibre="http://calibre.kovidgoyal.net/2009/metadata"> <dc:title>Operation Cameo</dc:title> <meta name="calibre:timestamp" content="2025-01-25T18:01:43.253715+00:00"/> <dc:language>en</dc:language> <dc:creator opf:file-as="alexwlchan" opf:role="aut">alexwlchan</dc:creator> <dc:identifier id="uuid_id" opf:scheme="uuid">13385d97-35a1-4e72-830b-9757916d38a7</dc:identifier> <meta name="calibre:title_sort" content="operation cameo"/> <dc:description><p>Some unusual orders arrive at Operation Mincemeat HQ.</p></dc:description> <dc:publisher>Archive of Our Own</dc:publisher> <dc:subject>Fanworks</dc:subject> <dc:subject>General Audiences</dc:subject> <dc:subject>Operation Mincemeat: A New Musical - SpitLip</dc:subject> <dc:subject>No Archive Warnings Apply</dc:subject> <dc:date>2023-12-14T00:00:00+00:00</dc:date> </metadata> … That dc: prefix was instantly familiar from my time working at Wellcome Collection – this is Dublin Core, a standard set of metadata fields used to describe books and other objects. I’m unsurprised to see it in an EPUB; this is exactly how I’d expect it to be used. I found an article that explains the structure of an EPUB file, which told me that I can find the content.opf file by looking at the root-path element inside the mandatory META-INF/container.xml file which is every EPUB. I wrote some code to find the content.opf file, then a few XPath expressions to extract the key fields, and I had the metadata I needed. Generate a cover image I sketched a simple cover design which shows the title and author. I wrote the first version of the drawing code in Pillow, because that’s what I’m familiar with. It was fine, but the code was quite flimsy – it didn’t wrap properly for long titles, and I couldn’t get custom fonts to work. Later I rewrote the app in JavaScript, so I had access to the HTML canvas element. This is another tool that I haven’t worked with before, so a fun chance to learn something new. The API felt fairly familiar, similar to other APIs I’ve used to build HTML elements. This time I did implement some line wrapping – there’s a measureText() API for canvas, so you can see how much space text will take up before you draw it. I break the text into words, and keeping adding words to a line until measureText tells me the line is going to overflow the page. I have lots of ideas for how I could improve the line wrapping, but it’s good enough for now. I was also able to get fonts working, so I picked Georgia to match the font used for titles on AO3. Here are some examples: I had several ideas for choosing the background colour. I’m trying to help my friend browse her collection of fic, and colour would be a useful way to distinguish things – so how do I use it? I realised I could get the fandom from the EPUB file, so I decided to use that. I use the fandom name as a seed to a random number generator, then I pick a random colour. This means that all the fics in the same fandom will get the same colour – for example, all the Star Wars stories are a shade of red, while Star Trek are a bluey-green. This was a bit harder than I expected, because it turns out that JavaScript doesn’t have a built-in seeded random number generator – I ended up using some snippets from a Stack Overflow answer, where bryc has written several pseudorandom number generators in plain JavaScript. I didn’t realise until later, but I designed something similar to the placeholder book covers in the Apple Books app. I don’t use Apple Books that often so it wasn’t a deliberate choice to mimic this style, but clearly it was somewhere in my subconscious. One difference is that Apple’s app seems to be picking from a small selection of background colours, whereas my code can pick a much nicer variety of colours. Apple’s choices will have been pre-approved by a designer to look good, but I think mine is more fun. Add the cover image to the EPUB My first attempt to add a cover image used pandoc: pandoc input.epub --output output.epub --epub-cover-image cover.jpeg This approach was no good: although it added the cover image, it destroyed the formatting in the rest of the EPUB. This made it easier to find the fic, but harder to read once you’d found it. An EPUB file I downloaded from AO3, before/after it was processed by pandoc. So I tried to do it myself, and it turned out to be quite easy! I unzipped another EPUB which already had a cover image. I found the cover image in OPS/images/cover.jpg, and then I looked for references to it in content.opf. I found two elements that referred to cover images: <?xml version="1.0" encoding="UTF-8"?> <package xmlns="http://www.idpf.org/2007/opf" version="3.0" unique-identifier="PrimaryID"> <metadata xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:opf="http://www.idpf.org/2007/opf"> <meta name="cover" content="cover-image"/> … </metadata> <manifest> <item id="cover-image" href="images/cover.jpg" media-type="image/jpeg" properties="cover-image"/> … </manifest> </package> This gave me the steps for adding a cover image to an EPUB file: add the image file to the zipped bundle, then add these two elements to the content.opf. Where am I going to deploy this? I wrote the initial prototype of this in Python, because that’s the language I’m most familiar with. Python has all the libraries I need: The zipfile module can unpack and modify the EPUB/ZIP The xml.etree or lxml modules can manipulate XML The Pillow library can generate images I built a small Flask web app: you upload the EPUB to my server, my server does some processing, and sends the EPUB back to you. But for such a simple app, do I need a server? I tried rebuilding it as a static web page, doing all the processing in client-side JavaScript. That’s simpler for me to host, and it doesn’t involve a round-trip to my server. That has lots of other benefits – it’s faster, less of a privacy risk, and doesn’t require a persistent connection. I love static websites, so can they do this? Yes! I just had to find a different set of libraries: The JSZip library can unpack and modify the EPUB/ZIP, and is the only third-party code I’m using in the tool Browsers include DOMParser for manipulating XML I’ve already mentioned the HTML <canvas> element for rendering the image This took a bit longer because I’m not as familiar with JavaScript, but I got it working. As a bonus, this makes the tool very portable. Everything is bundled into a single HTML file, so if you download that file, you have the whole tool. If my friend finds this tool useful, she can save the file and keep a local copy of it – she doesn’t have to rely on my website to keep using it. What should it look like? My first design was very “engineer brain” – I just put the basic controls on the page. It was fine, but it wasn’t good. That might be okay, because the only person I need to be able to use this app is my friend – but wouldn’t it be nice if other people were able to use it? If they’re going to do that, they need to know what it is – most people aren’t going to read a 2,500 word blog post to understand a tool they’ve never heard of. (Although if you have read this far, I appreciate you!) I started designing a proper page, including some explanations and descriptions of what the tool is doing. I got something that felt pretty good, including FAQs and acknowledgements, and I added a grey area for the part where you actually upload and download your EPUBs, to draw the user’s eye and make it clear this is the important stuff. But even with that design, something was missing. I realised I was telling you I’d create covers, but not showing you what they’d look like. Aha! I sat down and made up a bunch of amusing titles for fanfic and fanfic authors, so now you see a sample of the covers before you upload your first EPUB: This makes it clearer what the app will do, and was a fun way to wrap up the project. What did I learn from this project? Don’t be scared of new file formats My first instinct was to look for a third-party library that could handle the “complexity” of EPUB files. In hindsight, I’m glad I didn’t find one – it forced me to learn more about how EPUBs work, and I realised I could write my own code using built-in libraries. EPUB files are essentially ZIP files, and I only had basic needs. I was able to write my own code. Because I didn’t rely on a library, now I know more about EPUBs, I have code that’s simpler and easier for me to understand, and I don’t have a dependency that may cause problems later. There are definitely some file formats where I need existing libraries (I’m not going to write my own JPEG parser, for example) – but I should be more open to writing my own code, and not jumping to add a dependency. Static websites can handle complex file manipulations I love static websites and I’ve used them for a lot of tasks, but mostly read-only display of information – not anything more complex or interactive. But modern JavaScript is very capable, and you can do a lot of things with it. Static pages aren’t just for static data. One of the first things I made that got popular was find untagged Tumblr posts, which was built as a static website because that’s all I knew how to build at the time. Somewhere in the intervening years, I forgot just how powerful static sites can be. I want to build more tools this way. Async JavaScript calls require careful handling The JSZip library I’m using has a lot of async functions, and this is my first time using async JavaScript. I got caught out several times, because I forgot to wait for async calls to finish properly. For example, I’m using canvas.toBlob to render the image, which is an async function. I wasn’t waiting for it to finish, and so the zip would be repackaged before the cover image was ready to add, and I got an EPUB with a missing image. Oops. I think I’ll always prefer the simplicity of synchronous code, but I’m sure I’ll get better at async JavaScript with practice. Final thoughts I know my friend will find this helpful, and that feels great. Writing software that’s designed for one person is my favourite software to write. It’s not hyper-scale, it won’t launch the next big startup, and it’s usually not breaking new technical ground – but it is useful. I can see how I’m making somebody’s life better, and isn’t that what computers are for? If other people like it, that’s a nice bonus, but I’m really thinking about that one person. Normally the one person I’m writing software for is me, so it’s extra nice when I can do it for somebody else. If you want to try this tool yourself, go to alexwlchan.net/my-tools/add-cover-to-ao3-epubs/ If you want to read the code, it’s all on GitHub. [If the formatting of this post looks odd in your feed reader, visit the original article]
I’ve been doing Dry January this year. One thing I missed was something for apéro hour, a beverage to mark the start of the evening. Something complex and maybe bitter, not like a drink you’d have with lunch. I found some good options. Ghia sodas are my favorite. Ghia is an NA apéritif based on grape juice but with enough bitterness (gentian) and sourness (yuzu) to be interesting. You can buy a bottle and mix it with soda yourself but I like the little cans with extra flavoring. The Ginger and the Sumac & Chili are both great. Another thing I like are low-sugar fancy soda pops. Not diet drinks, they still have a little sugar, but typically 50 calories a can. De La Calle Tepache is my favorite. Fermented pineapple is delicious and they have some fun flavors. Culture Pop is also good. A friend gave me the Zero book, a drinks cookbook from the fancy restaurant Alinea. This book is a little aspirational but the recipes are doable, it’s just a lot of labor. Very fancy high end drink mixing, really beautiful flavor ideas. The only thing I made was their gin substitute (mostly junipers extracted in glycerin) and it was too sweet for me. Need to find the right use for it, a martini definitely ain’t it. An easier homemade drink is this Nonalcoholic Dirty Lemon Tonic. It’s basically a lemonade heavily flavored with salted preserved lemons, then mixed with tonic. I love the complexity and freshness of this drink and enjoy it on its own merits. Finally, non-alcoholic beer has gotten a lot better in the last few years thanks to manufacturing innovations. I’ve been enjoying NA Black Butte Porter, Stella Artois 0.0, Heineken 0.0. They basically all taste just like their alcoholic uncles, no compromise. One thing to note about non-alcoholic substitutes is they are not cheap. They’ve become a big high end business. Expect to pay the same for an NA drink as one with alcohol even though they aren’t taxed nearly as much.
The first time we had to evacuate Malibu this season was during the Franklin fire in early December. We went to bed with our bags packed, thinking they'd probably get it under control. But by 2am, the roaring blades of fire choppers shaking the house got us up. As we sped down the canyon towards Pacific Coast Highway (PCH), the fire had reached the ridge across from ours, and flames were blazing large out the car windows. It felt like we had left the evacuation a little too late, but they eventually did get Franklin under control before it reached us. Humans have a strange relationship with risk and disasters. We're so prone to wishful thinking and bad pattern matching. I remember people being shocked when the flames jumped the PCH during the Woolsey fire in 2017. IT HAD NEVER DONE THAT! So several friends of ours had to suddenly escape a nightmare scenario, driving through burning streets, in heavy smoke, with literally their lives on the line. Because the past had failed to predict the future. I feel into that same trap for a moment with the dramatic proclamations of wind and fire weather in the days leading up to January 7. Warning after warning of "extremely dangerous, life-threatening wind" coming from the City of Malibu, and that overly-bureaucratic-but-still-ominous "Particularly Dangerous Situation" designation. Because, really, how much worse could it be? Turns out, a lot. It was a little before noon on the 7th when we first saw the big plumes of smoke rise from the Palisades fire. And immediately the pattern matching ran astray. Oh, it's probably just like Franklin. It's not big yet, they'll get it out. They usually do. Well, they didn't. By the late afternoon, we had once more packed our bags, and by then it was also clear that things actually were different this time. Different worse. Different enough that even Santa Monica didn't feel like it was assured to be safe. So we headed far North, to be sure that we wouldn't have to evacuate again. Turned out to be a good move. Because by now, into the evening, few people in the connected world hadn't started to see the catastrophic images emerging from the Palisades and Eaton fires. Well over 10,000 houses would ultimately burn. Entire neighborhoods leveled. Pictures that could be mistaken for World War II. Utter and complete destruction. By the night of the 7th, the fire reached our canyon, and it tore through the chaparral and brush that'd been building since the last big fire that area saw in 1993. Out of some 150 houses in our immediate vicinity, nearly a hundred burned to the ground. Including the first house we moved to in Malibu back in 2009. But thankfully not ours. That's of course a huge relief. This was and is our Malibu Dream House. The site of that gorgeous home office I'm so fond to share views from. Our home. But a house left standing in a disaster zone is still a disaster. The flames reached all the way up to the base of our construction, incinerated much of our landscaping, and devoured the power poles around it to dysfunction. We have burnt-out buildings every which way the eye looks. The national guard is still stationed at road blocks on the access roads. Utility workers are tearing down the entire power grid to rebuild it from scratch. It's going to be a long time before this is comfortably habitable again. So we left. That in itself feels like defeat. There's an urge to stay put, and to help, in whatever helpless ways you can. But with three school-age children who've already missed over a months worth of learning from power outages, fire threats, actual fires, and now mudslide dangers, it was time to go. None of this came as a surprise, mind you. After Woolsey in 2017, Malibu life always felt like living on borrowed time to us. We knew it, even accepted it. Beautiful enough to be worth the risk, we said. But even if it wasn't a surprise, it's still a shock. The sheer devastation, especially in the Palisades, went far beyond our normal range of comprehension. Bounded, as it always is, by past experiences. Thus, we find ourselves back in Copenhagen. A safe haven for calamities of all sorts. We lived here for three years during the pandemic, so it just made sense to use it for refuge once more. The kids' old international school accepted them right back in, and past friendships were quickly rebooted. I don't know how long it's going to be this time. And that's an odd feeling to have, just as America has been turning a corner, and just as the optimism is back in so many areas. Of the twenty years I've spent in America, this feels like the most exciting time to be part of the exceptionalism that the US of A offers. And of course we still are. I'll still be in the US all the time on both business, racing, and family trips. But it won't be exclusively so for a while, and it won't be from our Malibu Dream House. And that burns.
Thou shalt not suffer a flaky test to live, because it’s annoying, counterproductive, and dangerous: one day it might fail for real, and you won’t notice. Here’s what to do.
The ware for January 2025 is shown below. Thanks to brimdavis for contributing this ware! …back in the day when you would get wares that had “blue wires” in them… One thing I wonder about this ware is…where are the ROMs? Perhaps I’ll find out soon! Happy year of the snake!
