A few days later

The previous section ended with two loose ends — series pages, and "decide what to do about rereads" — and an accounting of the process. It has been two days, and the accounting needs redoing.

The flood

The WordPress site was "on and off", but of course the off years didn't stop me reading, they only stopped me logging. Or, actually: they stopped me logging there. The books went to Goodreads instead, or got a review on my other weblog, or nowhere at all. As long as adding a book meant dealing with that WordPress install, none of it was ever coming back.

Weellll... it kind of came back in two days. A Goodreads export, compared against the catalogue, turned up some 340 dated reads that had simply never made it onto the site; a small interactive importer (look at the metadata, pick a cover from candidates keyed to the actual ISBN, next) walked through the lot. Then the series gaps — books I demonstrably read because I read the sequel — another 43. Then recent audiobooks from my Audible account: 29, of which 9 turned out to be rereads of things already on the site.

Net result: 416 new book pages. The catalogue went from 935 posts to 1,351.

The reviews that lived next door

The backfilled books mostly arrived without reviews — but a lot of the reviews existed, on my personal weblog, which for years cheerfully carried "Gelezen: …" posts that (almost) never got cross-posted to the booklog. Same story as the authors: a content-based matcher, strict auto-apply for the beyond-doubt cases, and a side-by-side decision screen for the rest — blog version, booklog version, or a mix, one keypress each. Around a hundred and thirty reviews recovered, plus 41 images restored to posts that had been quietly serving broken links since who knows when.

Rereads: decided

Every read stays its own post — own date, own rating, own review — and a later read carries a rereadOf pointing at the first one. Explicitly linked, not matched by title, because I own two different books called Nemesis and one book with two different titles. Detail pages cross-link the readings ("Herlezen op …"); series pages collapse the reads of one book into a single tile that lists every date (and has fun overlapping covers, and yes, I know those aren't ideal on mobile). That also means the series pages got their promised attention, so both loose ends are tied off.

The small stuff

A pile of little interface things too, none of which merits its own heading, all of which would have stayed on the someday-list forever on the old site. The lists got real page numbers — first and last page always visible, an ellipsis bridging the middle — and the keyboard pages along with them: ← and → for newer and older, Home and End for the ends. The ← / → arrows on a book page itself now follow the direction of the list you're browsing: the gelezen shelf runs newest-first, so → means "older"; the te-lezen shelf runs oldest-first, so → means "the one I added after it". (Arrows that always meant newer/older regardless of which way the list was pointing turned out to be exactly as disorienting as that sounds.) And the reading-progress veil — the cover that uncovers itself as I get further into a book — used to live only on the aan-het-lezen shelf; it now also shows on series pages and on the book's own page, and a book I haven't started sits under a full veil. Small stuff, like I said. But it's the small stuff that makes a site feel like someone lives there.

The app

There is also, now, an app. Adding books by talking to an agent in a code editor is a perfectly fine desk experience and a terrible couch one, and books get finished on the couch. So content management moved into a small mobile app that simply does what I would do by hand — edit the Markdown, run the checks, commit, push, let the build pipeline do the rest — while the site itself stays exactly as static as it was. That one got its own write-up, in a register I will not be apologising for.

And a stats page

Because of course there is a stats page now: books per year, categories, page-count histograms, languages — computed at build time, drawn as static SVG, no database, no API, nothing to hack. I'll get back to this later, I pretty sure, because it really is too basic right now.

The moral of the strangler fig was parity first, improvement second. What the fig doesn't tell you is what happens after: once the thing is pleasant to touch, you touch it. Twenty-odd years of deferred logging cleared in two days — not because two days is what it takes, but because it was never about the time.

(Reading continued throughout, I should note. Victorian Psycho advanced sixty-five pages while all this was going on. Priorities intact.)

And on, and on, and on

When the cost of rework is just the cost of explaining to an AI exactly what you want, rework never stops. After finishing Victorian Psycho I had Claude do the following:

1. Make Hugo generate cover webp thumbnails for the boosk in lists (turns out it used full-size images).
2. In "te lezen", show only the covers.
3. Make the current "Gelezen" page the home page and change the title to "Recent gelezen"; make the "gelezen" link in the main navigation go to an archive page. Navigating decades of books read with just previous/next and search was not the best experience. List all the years (descending) with the number of books; clicking on a year goes to a separate page per year with the books listed. For high-volume years, place a compact, horizontal row of month anchor links (Jan · Feb · Mar...) at the top of the page that jumps down to that month's section (only list the months that have books). For low-volume years, omit that row of anchor links. The user just scrolls through the short list.
4. Changed my mind about Home/End navigation: breaking in-page navigation to have it navigate the list was just confusing. A small affordance concession, too: style the right arrow / left arrow on the lists and detail pages as a small keyboard navigation button.
5. Make the treemap half the height, add hover labels to the smaller graphs. On desktop, instead of showing 4 small graphs, show 2 on a line. Drop the "zes reeksen" -- that doesn't really work.
6. Add a title attribute tooltip to covers in the lists; make the date a shade lighter for the sake of WCAG AA-ness.