More from Dan Quach Blog
AI Updates There is a lot of chatter about 2025 being the year of agentic frameworks. To me, this means a system in which a subset can allow AI models to take independent actions based on their environment, typically interacting with external APIs or interfaces. The terminology around this concept is still evolving, and definitions […]
Prompt Engineering – Meta Analysis Whitepaper One of my favorite AI podcasts, Latent Space, recently featured Sander Schulhoff, one of the authors of a comprehensive research paper on prompt engineering. This meta-study reviews over 1,600 published papers, with co-authors from OpenAI, Microsoft, and Stanford. [podcast] https://www.latent.space/p/learn-prompting [whitepaper] https://arxiv.org/abs/2406.06608 The whitepaper is an interesting academic deep dive into prompting, […]
Growing up I had the same dentist from childhood to adulthood. My dentist’s office was run by Dentist Chung (in Vietnamese I called him Bác Sĩ Chung – which means Dr Chung translated directly) and his sister running the office. The office was in Garden Grove, in between the Korean and Vietnamese districts. Walking in […]
Data Engineering and AIChip Huyen, who came out of Stanford and is active in the AI space recently wrote an article on what she learned by looking at the 900 most popular open source AI tools. https://huyenchip.com/2024/03/14/ai-oss.html In data engineering, one of our primary usages of AI is really just prompt engineering. Use Case 1: Data Migration Before LLMs, when we […]
More in programming
No newsletter next week, I'm teaching a TLA+ workshop. Speaking of which: I spend a lot of time thinking about formal methods (and TLA+ specifically) because it's where the source of almost all my revenue. But I don't share most of the details because 90% of my readers don't use FM and never will. I think it's more interesting to talk about ideas from FM that would be useful to people outside that field. For example, the idea of "property strength" translates to the idea that some tests are stronger than others. Another possible export is how FM approaches nondeterminism. A nondeterministic algorithm is one that, from the same starting conditions, has multiple possible outputs. This is nondeterministic: # Pseudocode def f() { return rand()+1; } When specifying systems, I may not encounter nondeterminism more often than in real systems, but I am definitely more aware of its presence. Modeling nondeterminism is a core part of formal specification. I mentally categorize nondeterminism into five buckets. Caveat, this is specifically about nondeterminism from the perspective of system modeling, not computer science as a whole. If I tried to include stuff on NFAs and amb operations this would be twice as long.1 1. True Randomness Programs that literally make calls to a random function and then use the results. This the simplest type of nondeterminism and one of the most ubiquitous. Most of the time, random isn't truly nondeterministic. Most of the time computer randomness is actually pseudorandom, meaning we seed a deterministic algorithm that behaves "randomly-enough" for some use. You could "lift" a nondeterministic random function into a deterministic one by adding a fixed seed to the starting state. # Python from random import random, seed def f(x): seed(x) return random() >>> f(3) 0.23796462709189137 >>> f(3) 0.23796462709189137 Often we don't do this because the point of randomness is to provide nondeterminism! We deliberately abstract out the starting state of the seed from our program, because it's easier to think about it as locally nondeterministic. (There's also "true" randomness, like using thermal noise as an entropy source, which I think are mainly used for cryptography and seeding PRNGs.) Most formal specification languages don't deal with randomness (though some deal with probability more broadly). Instead, we treat it as a nondeterministic choice: # software if rand > 0.001 then return a else crash # specification either return a or crash This is because we're looking at worst-case scenarios, so it doesn't matter if crash happens 50% of the time or 0.0001% of the time, it's still possible. 2. Concurrency # Pseudocode global x = 1, y = 0; def thread1() { x++; x++; x++; } def thread2() { y := x; } If thread1() and thread2() run sequentially, then (assuming the sequence is fixed) the final value of y is deterministic. If the two functions are started and run simultaneously, then depending on when thread2 executes y can be 1, 2, 3, or 4. Both functions are locally sequential, but running them concurrently leads to global nondeterminism. Concurrency is arguably the most dramatic source of nondeterminism. Small amounts of concurrency lead to huge explosions in the state space. We have words for the specific kinds of nondeterminism caused by concurrency, like "race condition" and "dirty write". Often we think about it as a separate topic from nondeterminism. To some extent it "overshadows" the other kinds: I have a much easier time teaching students about concurrency in models than nondeterminism in models. Many formal specification languages have special syntax/machinery for the concurrent aspects of a system, and generic syntax for other kinds of nondeterminism. In P that's choose. Others don't special-case concurrency, instead representing as it as nondeterministic choices by a global coordinator. This more flexible but also more inconvenient, as you have to implement process-local sequencing code yourself. 3. User Input One of the most famous and influential programming books is The C Programming Language by Kernighan and Ritchie. The first example of a nondeterministic program appears on page 14: For the newsletter readers who get text only emails,2 here's the program: #include /* copy input to output; 1st version */ main() { int c; c = getchar(); while (c != EOF) { putchar(c); c = getchar(); } } Yup, that's nondeterministic. Because the user can enter any string, any call of main() could have any output, meaning the number of possible outcomes is infinity. Okay that seems a little cheap, and I think it's because we tend to think of determinism in terms of how the user experiences the program. Yes, main() has an infinite number of user inputs, but for each input the user will experience only one possible output. It starts to feel more nondeterministic when modeling a long-standing system that's reacting to user input, for example a server that runs a script whenever the user uploads a file. This can be modeled with nondeterminism and concurrency: We have one execution that's the system, and one nondeterministic execution that represents the effects of our user. (One intrusive thought I sometimes have: any "yes/no" dialogue actually has three outcomes: yes, no, or the user getting up and walking away without picking a choice, permanently stalling the execution.) 4. External forces The more general version of "user input": anything where either 1) some part of the execution outcome depends on retrieving external information, or 2) the external world can change some state outside of your system. I call the distinction between internal and external components of the system the world and the machine. Simple examples: code that at some point reads an external temperature sensor. Unrelated code running on a system which quits programs if it gets too hot. API requests to a third party vendor. Code processing files but users can delete files before the script gets to them. Like with PRNGs, some of these cases don't have to be nondeterministic; we can argue that "the temperature" should be a virtual input into the function. Like with PRNGs, we treat it as nondeterministic because it's useful to think in that way. Also, what if the temperature changes between starting a function and reading it? External forces are also a source of nondeterminism as uncertainty. Measurements in the real world often comes with errors, so repeating a measurement twice can give two different answers. Sometimes operations fail for no discernable reason, or for a non-programmatic reason (like something physically blocks the sensor). All of these situations can be modeled in the same way as user input: a concurrent execution making nondeterministic choices. 5. Abstraction This is where nondeterminism in system models and in "real software" differ the most. I said earlier that pseudorandomness is arguably deterministic, but we abstract it into nondeterminism. More generally, nondeterminism hides implementation details of deterministic processes. In one consulting project, we had a machine that received a message, parsed a lot of data from the message, went into a complicated workflow, and then entered one of three states. The final state was totally deterministic on the content of the message, but the actual process of determining that final state took tons and tons of code. None of that mattered at the scope we were modeling, so we abstracted it all away: "on receiving message, nondeterministically enter state A, B, or C." Doing this makes the system easier to model. It also makes the model more sensitive to possible errors. What if the workflow is bugged and sends us to the wrong state? That's already covered by the nondeterministic choice! Nondeterministic abstraction gives us the potential to pick the worst-case scenario for our system, so we can prove it's robust even under those conditions. I know I beat the "nondeterminism as abstraction" drum a whole lot but that's because it's the insight from formal methods I personally value the most, that nondeterminism is a powerful tool to simplify reasoning about things. You can see the same approach in how I approach modeling users and external forces: complex realities black-boxed and simplified into nondeterministic forces on the system. Anyway, I hope this collection of ideas I got from formal methods are useful to my broader readership. Lemme know if it somehow helps you out! I realized after writing this that I already talked wrote an essay about nondeterminism in formal specification just under a year ago. I hope this one covers enough new ground to be interesting! ↩ There is a surprising number of you. ↩
Trump is doing Europe a favor by revealing the true cost of its impotency. Because, in many ways, he has the manners and the honesty of a child. A kid will just blurt out in the supermarket "why is that lady so fat, mommy?". That's not a polite thing to ask within earshot of said lady, but it might well be a fair question and a true observation! Trump is just as blunt when he essentially asks: "Why is Europe so weak?". Because Europe is weak, spiritually and militarily, in the face of Russia. It's that inherent weakness that's breeding the delusion that Russia is at once both on its last legs, about to lose the war against Ukraine any second now, and also the all-potent superpower that could take over all of Europe, if we don't start World Word III to counter it. This is not a coherent position. If you want peace, you must be strong. The big cats in the international jungle don't stick to a rules-based order purely out of higher principles, but out of self-preservation. And they can smell weakness like a tiger smells blood. This goes for Europe too. All too happy to lecture weaker countries they do not fear on high-minded ideals of democracy and free speech, while standing aghast and weeping powerlessly when someone stronger returns the favor. I'm not saying that this is right, in some abstract moral sense. I like the idea of a rules-based order. I like the idea of territorial sovereignty. I even like the idea that the normal exchanges between countries isn't as blunt and honest as those of a child in the supermarket. But what I like and "what is" need separating. Europe simply can't have it both ways. Be weak militarily, utterly dependent on an American security guarantee, and also expect a seat at the big-cat table. These positions are incompatible. You either get your peace dividend -- and the freedom to squander it on net-zero nonsense -- or you get to have a say in how the world around you is organized. Which brings us back to Trump doing Europe a favor. For all his bluster and bullying, America is still a benign force in its relation to Europe. We're being punked by someone from our own alliance. That's a cheap way of learning the lesson that weakness, impotence, and peace-dividend thinking is a short-term strategy. Russia could teach Europe a far more costly lesson. So too China. All that to say is that Europe must heed the rude awakening from our cowboy friends across the Atlantic. They may be crude, they may be curt, but by golly, they do have a point. Get jacked, Europe, and you'll no longer get punked. Stay feeble, Europe, and the indignities won't stop with being snubbed in Saudi Arabia.
Last year I wrote about using static websites for tiny archives. The idea is that I create tiny websites to store and describe my digital collections. There are several reasons I like this approach: HTML is flexible and lets me display data in a variety of ways; it’s likely to remain readable for a long time; it lets me add more context than a folder full of files. I’m converting more and more of my local data to be stored in static websites – paperwork I’ve scanned, screenshots I’ve taken, and web pages I’ve bookmarked. I really like this approach. I got a lot of positive feedback, but the most common reply was “please share some source code”. People wanted to see examples of the HTML and JavaScript I was using I deliberately omitted any code from the original post, because I wanted to focus on the concept, not the detail. I was trying to persuade you that static websites are a good idea for storing small archives and data sets, and I didn’t want to get distracted by the implementation. There’s also no single code base I could share – every site I build is different, and the code is often scrappy or poorly documented. I’ve built dozens of small sites this way, and there’s no site that serves as a good example of this approach – they’re all built differently, implement a subset of my ideas, or have hard-coded details. Even if I shared some source code, it would be difficult to read or understand what’s going on. However, there’s clearly an appetite for that sort of explanation, so this follow-up post will discuss the “how” rather than the “why”. There’s a lot of code, especially JavaScript, which I’ll explain in small digestible snippets. That’s another reason I didn’t describe this in the original post – I didn’t want anyone to feel overwhelmed or put off. A lot of what I’m describing here is nice-to-have, not essential. You can get started with something pretty simple. I’ll go through a feature at a time, as if we were building a new static site. I’ll use bookmarks as an example, but there’s nothing in this post that’s specific to bookmarking. If you’d like to see everything working together, check out the demo site. It includes the full code for all the sections in this post. Let’s dive in! Start with a hand-written HTML page (demo) Reduce repetition with JavaScript templates (demo) Add filtering to find specific items (demo) Introduce sorting to bring order to your data (demo) Use pagination to break up long lists (demo) Provide feedback with loading states and error handling (demo 1, demo 2) Test the code with QUnit and Playwright Manipulate the metadata with Python Store the website code in Git Closing thoughts demo) A website can be a single HTML file you edit by hand. Open a text editor like TextEdit or Notepad, copy-paste the following text, and save it in a file named bookmarks.html. <h1>Bookmarks</h1> <ul> <li><a href="https://estherschindler.medium.com/the-old-family-photos-project-lessons-in-creating-family-photos-that-people-want-to-keep-ea3909129943">Lessons in creating family photos that people want to keep, by Esther Schindler (2018)</a></li> <li><a href="https://www.theatlantic.com/technology/archive/2015/01/why-i-am-not-a-maker/384767/">Why I Am Not a Maker, by Debbie Chachra (The Atlantic, 2015)</a></li> <li><a href="https://meyerweb.com/eric/thoughts/2014/06/10/so-many-nevers/">So Many Nevers, by Eric Meyer (2014)</a></li> </ul> If you open this file in your web browser, you’ll see a list of three links. You can also check out my demo page to see this in action. This is an excellent way to build a website. If you stop here, you’ve got all the flexibility and portability of HTML, and this file will remain readable for a very long time. I build a lot of sites this way. I like it for small data sets that I know are never going to change, or which change very slowly. It’s simple, future-proof, and easy to edit if I ever need to. demo) As you store more data, it gets a bit tedious to keep copying the HTML markup for each item. Wouldn’t it be useful if we could push it into a reusable template? When a site gets bigger, I convert the metadata into JSON, then I use JavaScript and template literals to render it on the page. Let’s start with a simple example of metadata in JSON. My real data has more fields, like date saved or a list of keyword tags, but this is enough to get the idea: const bookmarks = [ { "url": "https://estherschindler.medium.com/the-old-family-photos-project-lessons-in-creating-family-photos-that-people-want-to-keep-ea3909129943", "title": "Lessons in creating family photos that people want to keep, by Esther Schindler (2018)" }, { "url": "https://www.theatlantic.com/technology/archive/2015/01/why-i-am-not-a-maker/384767/", "title": "Why I Am Not a Maker, by Debbie Chachra (The Atlantic, 2015)" }, { "url": "https://meyerweb.com/eric/thoughts/2014/06/10/so-many-nevers/", "title": "So Many Nevers, by Eric Meyer (2014)" } ]; Then I have a function that renders the data for a single bookmark as HTML: function Bookmark(bookmark) { return ` <li> <a href="${bookmark.url}">${bookmark.title}</a> </li> `; } Having a function that returns HTML is inspired by React and Next.js, where code is split into “components” that each render part of the web app. This function is simpler than what you’d get in React. Part of React’s behaviour is that it will re-render the page if the data changes, but my function won’t do that. That’s okay, because my data isn’t going to change. The HTML gets rendered once when the page loads, and that’s enough. I’m using a template literal because I find it simple and readable. It looks pretty close to the actual HTML, so I have a pretty good idea of what’s going to appear on the page. Template literals are dangerous if you’re getting data from an untrusted source – it could allow somebody to inject arbitrary HTML into your page – but I’m writing all my metadata, so I trust it. I know there are other ways to construct HTML in JavaScript, like document.createElement(), the <template> element, or Web Components – but template literals have always been sufficient for me, and I’ve never had a reason to explore other options. Now we have to call this function when the page loads, and render the list of bookmarks. Here’s the rest of the code: <script> window.addEventListener("DOMContentLoaded", () => { document.querySelector("#listOfBookmarks").innerHTML = bookmarks.map(Bookmark).join(""); }); </script> <h1>Bookmarks</h1> <ul id="listOfBookmarks"></ul> I’m listening for the DOMContentLoaded event, which occurs when the HTML page has been fully parsed. When that event occurs, it looks for <ul id="listOfBookmarks"> in the page, and inserts the HTML for the list of bookmarks. We have to wait for this event so the <ul> actually exists. If we tried to run it immediately, it might run before the <ul> exists, and then it wouldn’t know where to insert the HTML. I’m using querySelector() to find the <ul> I want to modify – this is a newer alternative to functions like getElementById(). It’s quite flexible, because I can target any CSS selector, and I find CSS rules easier to remember than the family of getElementBy* functions. Although it’s slightly slower in benchmarks, the difference is negligible and it’s easier for me to remember. If you want to see this page working, check out the demo page. I use this pattern as a starting point for a lot of my static sites – metadata in JSON, some functions that render HTML, and an event listener that renders the whole page after it loads. Once I have the basic site, I add data, render more HTML, and write CSS styles to make it look pretty. This is where I can have fun, and really customise each site. I keep tweaking until I have something I like. I’m ignoring CSS because that could be a whole other post, and there’s a vintage charm to unstyled HTML – it’s fine for what we’re discussing today. What else can we do? demo) As the list gets even longer, it’s useful to have a way to find specific items in the list – I don’t want to scroll the whole thing every time. I like adding keyword tags to my data, and then filtering for items with particular tags. If I add other metadata fields, I could filter on those too. Here’s a brief sketch of the sort of interface I like: I like to be able to define a series of filters, and apply them to focus on a specific subset of items. I like to combine multiple filters to refine my search, and to see a list of applied filters with a way to remove them, if I’ve filtered too far. I like to apply filters from a global menu, or to use controls on each item to find similar items. I use URL query parameters to store the list of currently-applied filters, for example: bookmarks.html?tag=animals&tag=wtf&publicationYear=2025 This means that any UI element that adds or removes a filter is a link to a new URL, so clicking it loads a new page, which triggers a complete re-render with the new filters. When I write filtering code, I try to make it as easy as possible to define new filters. Every site needs a slightly different set of filters, but the overall principle is always the same: here’s a long list of items, find the items that match these rules. Let’s start by expanding our data model to include a couple of new fields: const bookmarks = [ { "url": "https://estherschindler.medium.com/the-old-family-photos-project-lessons-in-creating-family-photos-that-people-want-to-keep-ea3909129943", "title": "Lessons in creating family photos that people want to keep, by Esther Schindler (2018)", "tags": ["photography", "preservation"], "publicationYear": "2018" }, … ]; Then we can define some filters we might use to narrow the list: const bookmarkFilters = [ { id: 'tag', label: 'tagged with', filterFn: (bookmark, tagName) => bookmark.tags.includes(tagName), }, { id: 'publicationYear', label: 'published in', filterFn: (bookmark, year) => bookmark.publicationYear === year, }, ]; Each filter has three fields: id matches the name of the associated URL query parameter label is how the filter will be described in the list of applied filters filterFn is a function that takes two arguments: a bookmark, and a filter value, and returns true/false depending on whether the bookmark matches this filter This list is the only place where I need to customise the filters for a particular site; the rest of the filtering code is completely generic. This means there’s only one place I need to make changes if I want to add or remove filters. The next piece of the filtering code is a generic function that filters a list of items, and takes the list of filters as an argument: /* * Filter a list of items. * * This function takes the list of items and available filters, and the * URL query parameters passed to the page. * * This function returns a list with the items that match these filters, * and a list of filters that have been applied. */ function filterItems({ items, filters, params }) { // By default, all items match, and no filters are applied. var matchingItems = items; var appliedFilters = []; // Go through the URL query params one by one, and look to // see if there's a matching filter. for (const [key, value] of params) { console.debug(`Checking query parameter ${key}`); const matchingFilter = filters.find(f => f.id === key); if (typeof matchingFilter === 'undefined') { continue; } // There's a matching filter! Go ahead and filter the // list of items to only those that match. console.debug(`Detected filter ${JSON.stringify(matchingFilter)}`); matchingItems = matchingItems.filter( item => matchingFilter.filterFn(item, value) ); // Construct a new query string that doesn't include // this filter. const altQuery = new URLSearchParams(params); altQuery.delete(key, value); const linkToRemove = "?" + altQuery.toString(); appliedFilters.push({ type: matchingFilter.id, label: matchingFilter.label, value, linkToRemove, }) } return { matchingItems, appliedFilters }; } This function doesn’t care what sort of items I’m passing, or what the actual filters are, so I can reuse it between different sites. It returns the list of matching items, and the list of applied filters. The latter allows me to show that list on the page. linkToRemove is a link to the same page with this filter removed, but keeping any other filters. This lets us provide a button that removes the filter. The final step is to wire this filtering into the page render. We need to make sure we only show items that match the filter, and show the user a list of applied filters. Here’s the new code: <script> window.addEventListener("DOMContentLoaded", () => { const params = new URLSearchParams(window.location.search); const { matchingItems: matchingBookmarks, appliedFilters } = filterItems({ items: bookmarks, filters: bookmarkFilters, params: params, }); document.querySelector("#appliedFilters").innerHTML = appliedFilters .map(f => `<li>${f.label}: ${f.value} <a href="${f.linkToRemove}">(remove)</a></li>`) .join(""); document.querySelector("#listOfBookmarks").innerHTML = matchingBookmarks.map(Bookmark).join(""); }); </script> <h1>Bookmarks</h1> <p>Applied filters:</p> <ul id="appliedFilters"></ul> <p>Bookmarks:</p> <ul id="listOfBookmarks"></ul> I stick to simple filters that can be phrased as a yes/no question, and I rely on my past self to have written sufficiently useful metadata. At least in static sites, I’ve never implemented anything like a fuzzy text search, where it’s less obvious whether a particular item should match. You can check out the filtering code on the demo page. demo) The next feature I usually implement is sorting. I build a dropdown menu with all the options, and picking one reloads the page with the new sort order. Here’s a quick design sketch: For example, I often sort by the date I saved an item, so I can find an item I saved recently. Another sort order I often use is “random”, which shuffles the items and is a fun way to explore the data. As with filters, I put the current sort order in a query parameter, for example: bookmarks.html?sortOrder=titleAtoZ As before, I want to write this in a generic way and share code between different sites. Let’s start by defining a list of sort options: const bookmarkSortOptions = [ { id: 'titleAtoZ', label: 'title (A to Z)', compareFn: (a, b) => a.title > b.title ? 1 : -1, }, { id: 'publicationYear', label: 'publication year (newest first)', compareFn: (a, b) => Number(b.publicationYear) - Number(a.publicationYear), }, ]; Each sort option has three fields: id is the value that will appear in the URL query parameter label is the human-readable label that will appear in the dropdown compareFn(a, b) is a function that compares two items, and will be passed directly to the JavaScript sort function. If it returns a negative value, then a sorts before b. If it returns a positve value, then a sorts after b. Next, we can define a function that will sort a list of items: /* * Sort a list of items. * * This function takes the list of items and available options, and the * URL query parameters passed to the page. * * It returns a list with the items in sorted order, and the * sort order that was applied. */ function sortItems({ items, sortOptions, params }) { // Did the user pass a sort order in the query parameters? const sortOrderId = getSortOrder(params); // What sort order are we using? // // Look for a matching sort option, or use the default if the sort // order is null/unrecognised. For now, use the first defined // sort order as the default. const defaultSort = sortOptions[0]; const selectedSort = sortOptions.find(s => s.id === sortOrderId) || defaultSort; console.debug(`Selected sort: ${JSON.stringify(selectedSort)}`); // Now apply the sort to the list of items. const sortedItems = items.sort(selectedSort.compareFn); return { sortedItems, appliedSortOrder: selectedSort }; } /* Get the current sort order from the URL query parameters. */ function getSortOrder(params) { return params.get("sortOrder"); } This function works with any list of items and sort orders, making it easy to reuse across different sites. I only have to define the list of sort orders once. This approach makes it easy to add new sort orders, and to write a component that renders a dropdown menu to pick the sort order: /* * Create a dropdown control to choose the sort order. When you pick * a different value, the page reloads with the new sort. */ function SortOrderDropdown({ sortOptions, appliedSortOrder }) { return ` <select onchange="setSortOrder(this.value)"> ${ sortOptions .map(({ id, label }) => ` <option value="${id}" ${id === appliedSortOrder.id ? 'selected' : ''}> ${label} </option> `) .join("") } </select> `; } function setSortOrder(sortOrderId) { const params = new URLSearchParams(window.location.search); params.set("sortOrder", sortOrderId); window.location.search = params.toString(); } Finally, we can wire the sorting code into the rest of the app. After filtering, we sort the items and then render the sorted list. We also show the sort controls on the page: <script> window.addEventListener("DOMContentLoaded", () => { const params = new URLSearchParams(window.location.search); const { matchingItems: matchingBookmarks, appliedFilters } = filterItems(…); … const { sortedItems: sortedBookmarks, appliedSortOrder } = sortItems({ items: matchingBookmarks, sortOptions: bookmarkSortOptions, params, }); document.querySelector("#sortOrder").innerHTML += SortOrderDropdown({ sortOptions: bookmarkSortOptions, appliedSortOrder }); document.querySelector("#listOfBookmarks").innerHTML = sortedBookmarks.map(Bookmark).join(""); }); </script> <p id="sortOrder">Sort by:</p> You can check out the sorting code on the demo page. demo) If you have a really long list of items, you may want to break them into multiple pages. This isn’t something I do very often. Modern web browsers are very performant, and you can put thousands of elements on the page without breaking a sweat. I’ve only had to add pagination in a couple of very image-heavy sites – if it’s a text-based site, I just show everything. (You may notice that, for example, there are no paginated lists anywhere on this site. By writing lean HTML, I can fit all my lists on a single page.) If I do want pagination, I stick to a classic design: As with other features, I use a URL query parameter to track the current page number: bookmarks.html?pageNumber=2 This code can be written in a completely generic way – it doesn’t have to care what sort of items we’re paginating. First, let’s write a function that will select a page of items for us. If we’re on page N, what items should we be showing? /* * Get a page of items. * * This function will reduce the list of items to the items that should * be shown on this particular page. */ function paginateItems({ items, pageNumber, pageSize }) { // Page numbers are 1-indexed, so page 1 corresponds to // the indices 0…(pageSize - 1). const startOfPage = (pageNumber - 1) * pageSize; const endOfPage = pageNumber * pageSize; const thisPage = items.slice(startOfPage, endOfPage); return { thisPage, totalPages: Math.ceil(items.length / pageSize), }; } In some of my sites, the page size is a suggestion rather than a hard rule. If there are 27 items and the page size is 25, I think it’s nicer to show all the items on one page than push a few items onto a second page which barely has anything on it. But that might reflect my general dislike of pagination, and it’s definitely a nice-to-have rather than a required feature. Once we know what page we’re on and how many pages there are, we can create a component to render some basic pagination controls: /* * Renders a list of pagination controls. * * This includes links to prev/next pages and the current page number. */ function PaginationControls({ pageNumber, totalPages, params }) { // If there are no pages, we don't need pagination controls. if (totalPages === 1) { return ""; } // Do we need a link to the previous page? Only if we're past page 1. if (pageNumber > 1) { const prevPageUrl = setPageNumber({ params, pageNumber: pageNumber - 1 }); prevPageLink = `<a href="${prevPageUrl}">← prev</a>`; } else { prevPageLink = null; } // Do we need a link to the next page? Only if we're before // the last page. if (pageNumber < totalPages) { const nextPageUrl = setPageNumber({ params, pageNumber: pageNumber + 1 }); nextPageLink = `<a href="${nextPageUrl}">next →</a>`; } else { nextPageLink = null; } const pageText = `Page ${pageNumber} of ${totalPages}`; // Construct the final result. return [prevPageLink, pageText, nextPageLink] .filter(p => p !== null) .join(" / "); } /* Returns a URL that points to the new page number. */ function setPageNumber({ params, pageNumber }) { const updatedParams = new URLSearchParams(params); updatedParams.set("pageNumber", pageNumber); return `?${updatedParams.toString()}`; } Finally, let’s wire this code into the rest of the app. We get the page number from the URL query parameters, paginate the list of filtered and sorted items, and show some pagination controls: <script> /* Get the current page number. */ function getPageNumber(params) { return Number(params.get("pageNumber")) || 1; } window.addEventListener("DOMContentLoaded", () => { const params = new URLSearchParams(window.location.search); const { matchingItems: matchingBookmarks, appliedFilters } = filterItems(…); const { sortedItems: sortedBookmarks, appliedSortOrder } = sortItems(…); const pageNumber = getPageNumber(params); const { thisPage: thisPageOfBookmarks, totalPages } = paginateItems({ items: sortedBookmarks, pageNumber, pageSize: 25, }); document.querySelector("#paginationControls").innerHTML += PaginationControls({ pageNumber, totalPages, params }); document.querySelector("#listOfBookmarks").innerHTML = thisPageOfBookmarks.map(Bookmark).join(""); }); </script> <p id="paginationControls">Pagination controls: </p> One thing that makes pagination a little tricky is that it affects filtering and sorting as well – when you change either of those, you probably want to reset to the first page. For example, if you’re filtering for animals and you’re on page 3, then you add a second filter for giraffes, you should reset to page 1. If you stay on page 3, it might be confusing if there are less than 3 pages of results with the new filter. The key to this is calling params.delete("pageNumber") when you update the URL query parameters. You can play with the pagination on the demo page. demo 1, demo 2) One problem with relying on JavaScript to render the page is that sometimes JavaScript goes wrong. For example, I write a lot of my metadata by hand, and a typo can create invalid JSON and break the page. There are also people who disable JavaScript, or sometimes it just doesn’t work. If I’m using the site, I can open the Developer Tools in my web browser and start debugging there – but that’s not a great experience. If you’re not expecting something to go wrong, it will just look like the page is taking a long time to load. We can do better. To start, we can add a <noscript> element that explains to users that they need to enable JavaScript. This will only be shown if they’ve disabled JavaScript: <noscript> <strong>You need to enable JavaScript to use this site!</strong> </noscript> I have a demo page which disables JavaScript, so you can see how the noscript tag behaves. This won’t help if JavaScript is broken rather than disabled, so we also need to add error handling. We can listen for the error event on the window, and report an error to the user – for example, if a script fails to load. <div id="errors"></div> <script> window.addEventListener("error", function(event) { document .querySelector('#errors') .innerHTML = `<strong>Something went wrong when loading the page!</strong>`; }); </script> We can also attach an onerror handler to specific script tags, which allows us to customise the error message – we can tell the user that a particular file failed to load. <script src="app.js" onerror="alert('Something went wrong while loading app.js')"></script> I have another demo page which has a basic error handler. Finally, I like to include a loading indicator, or some placeholder text that will be replaced when the page will finish loading – this tells the user where they can expect to see something load in. <ul id="listOfBookmarks">Loading…</ul> It’s somewhat rare for me to add a loading indicator or error handling, just because I’m the only user of my static sites, and it’s easier for me to use the developer tools when something breaks. But providing mechanisms for the user to understand what’s going on is crucial if you want to build static sites like this that other people will use. Test the code with QUnit and Playwright If I’m writing a very complicated viewer, it’s helpful to have tests. I’ve found two test frameworks that I particularly like for this purpose. QUnit is a JavaScript library that I use for unit testing – to me, that means testing individual functions and components. For example, QUnit was very helpful when I was writing the early iterations of the sorting and filtering code, and writing tests caught a number of mistakes. You can run QUnit in the browser, and it only requires two files, so I can test a project without creating a whole JavaScript build system or dependency tree. Here’s an example of a QUnit test: QUnit.test("sorts bookmarks by title", function(assert) { // Create three bookmarks with different titles const bookmarkA = { title: "Almanac for apples" }; const bookmarkC = { title: "Compendium of coconuts" }; const bookmarkP = { title: "Page about papayas" }; const params = new URLSearchParams("sortOrder=titleAtoZ"); // Pass the bookmarks in the wrong order, so they can't be sorted // correctly "by accident" const { sortedItems, appliedSortOrder } = sortItems({ items: [bookmarkC, bookmarkA, bookmarkP], sortOptions: bookmarkSortOptions, params, }); // Check the bookmarks have been sorted in the right order assert.deepEqual(sortedItems, [bookmarkA, bookmarkC, bookmarkP]); }); You can see this test running in the browser in my demo page. Playwright is a testing library that can open a web app in a real web browser, interact with the page, and check that the app behaves correctly. It’s often used for dynamic web apps, but it works just as well for static pages. For example, it can test that if you select a new sort order, the page reloads and show results in the correct order. Here’s an example of a simple test written with Playwright in Python: from playwright.sync_api import expect, sync_playwright with sync_playwright() as p: browser = p.webkit.launch() # Open the HTML file in the browser page = browser.new_page() page.goto('file:///Users/alexwlchan/Sites/sorting.html') # Look for an <li> element with one of the bookmarks -- this will # only appear if the page has rendered correctly. expect(page.get_by_text("So Many Nevers")).to_be_visible() browser.close() These tools are a great safety net for catching mistakes, but I don’t always need them. I only write tests for my more complicated sites – when the sorting/filtering code is particularly complex, there’s a lot of rendering code, or I anticipate making major changes in future. I don’t bother with tests when the site is simple and unlikely to change, and I can just do manual checks when I write it the first time. Tests are less useful if I know I’ll never make changes. This is getting away from the idea of a self-contained static website, because now I’m relying on third-party code, and for Playwright I need to maintain a working Python environment. I’m okay with this, because the website is still usable even if I can no longer run the tests. These are useful sidecar tools, but I only need them if I’m making changes. If I finish a site and I know I won’t change it again, I don’t need to worry about whether the tests will still work years later. Manipulate the metadata with Python For small sites, we could write all this JavaScript directly in <script> tags or in a single file. As we get more data, splitting the metadata and application logic makes everything easier to manage. One pattern I’ve adopted is to put all the item metadata into a single, standalone JavaScript file that assigns a single variable: const bookmarks = […]; and then load that file in the HTML page with a <script src="metadata.js"> element. I use JavaScript rather than pure JSON because browsers don’t allow fetching local JSON files via file://. If you open an HTML page without a web server, the browser will block requests to fetch a JSON file because of security restrictions. By storing data in a JavaScript file instead, I can load it with a simple <script> tag. I wrote a small Python library javascript-data-files that lets me interact with JSON stored this way. This allows me to write scripts that add data to the metadata file (like saving a new bookmark) or to verify the existing metadata (like checking that I have an archived copy of every bookmark). I’ll write more about this in future posts, because this one is long enough already. For example, let’s add a new bookmark to the metadata.js file: from javascript_data_files import read_js, write_js bookmarks = read_js("metadata.js", varname="bookmarks") bookmarks.append({ "url": "https://www.theguardian.com/lifeandstyle/2019/jan/13/ella-risbridger-john-underwood-friendship-life-new-family", "title": "When my world fell apart, my friends became my family, by Ella Risbridger (2019)" }) write_js("metadata.js", varname="bookmarks", value=bookmarks) We’re starting to blur the line between a static site and a static site generator. These scripts only work if I have a working Python environment, which is less future-proof than pure HTML. I’m happy with this compromise, because the website is fully functional without them – I only need to run these scripts if I’m modifying the metadata. If I stop making changes and the Python environment breaks, I can still read everything I’ve already saved. Store the website code in Git I create Git repositories for all of my local websites. This allows me to track changes, and it means I can experiment freely – I can always roll back if I break something. These Git repositories only live on my local machine. I run git init . in the folder, I create commits to record any changes, and that’s it. I don’t push the repository to GitHub or another remote Git server. (Although I do have backups of every site, of course.) Git has a lot of features for writing code in a collaborative environment, but I don’t need any of those here – I’m the only person working on these sites. Most of the time, I just use two commands: $ git add bookmarks.html $ git commit -m "Add filtering by author" This creates a labelled snapshot of my latest changes to bookmarks.html. I only track the text files in Git – the HTML, CSS, and JavaScript. I don’t track binary files like images and videos. Git struggles with those larger files, and I don’t edit those as much as the text files, so having them in version control is less useful. I write a gitignore file to ignore all of them. Closing thoughts There are lots of ideas here, but you don’t need to use all of them – most of my sites only use a few. Every site is different, and you can pick what makes most sense for your project. If you’re building a static site for a tiny archive, start with a simple HTML file. Add features like templates, sorting, and filtering incrementally as they become useful. You don’t need to add them all upfront – that can make things more complicated than they need to be. This approach can scale from simple collections to sophisticated archives. A static website built with HTML and JavaScript is easy to maintain and modify, has no external dependencies, and is future-proof against a lot of technological changes. I’ve come to love using static websites to store my local data. They’re flexible, resilient, and surprisingly powerful. I hope you’ll consider it too, and that these ideas help you get started. [If the formatting of this post looks odd in your feed reader, visit the original article]
Intellectual property is a really dumb idea. “But piracy is theft. Clean and simple. It’s smash and grab. It ain’t no different than smashing a window at Tiffany’s and grabbing merchandise.” - Joe Biden, 46th president of the USA Except it isn’t and Joe Biden is a senile moron. Because when you smash the windows and grab the stuff, Tiffany’s no longer has the stuff. With piracy, everyone has the stuff. It’s a lot more like taking a picture, which Tiffany’s probably encourages. Win-win cooperation. Wealth is being increasingly concentrated. What’s shocking to me is how much everyone still cares about money. Even the die-hard complain about capitalism type deeply cares, because the opposite of love isn’t hate, it’s indifference. I hate scammers, but I’m pretty indifferent to money. The best outcome of AI is if it delivers huge amounts of value to society but no profit to anyone. The old days of the Internet were this goldmine. The Internet delivered huge value but no profit, and that’s why it was good. Suddenly we had all these new powers. Then people figured out how to monetize it. It was a race to extract every tiny bit of value, and now we have today’s Internet. Can this play out differently with AI? Let’s build technology and open source software that market breaks everything. Let’s demoralize the scammers so hard that they don’t even try. Every loser and grifter will be gone from technology because there’s nothing to be gained there. They can play golf all day or something. If I ever figure out how to channel power like Elon, I will do this. Spin up open source projects in every sector to eliminate all the capturable value. This is what I’m trying to do with comma.ai and tinygrad. I dream of a day when company valuations halve when I create a GitHub repo. Someday.
A personal guide to the most useful books for understanding operating systems