İsmet Güzelgün

Leaving MyAnimeList

· 2 minute read

I’ve been on MyAnimeList since 2014. It tracks well enough, the community is fine, and inertia is strong — none of which mattered the moment I wanted to actually do something with my data.

MAL has no public API for users. Not a deprecated one, not a rate-limited one — none. If you want your own list out, you can ask the site for an XML export. That’s it.

AniList, on the other hand, ships a public GraphQL API. You can read your list, write to it, query everything down to per-episode metadata. So I switched, and I built anilist-user-analytics to play with the data — completion patterns, score distributions, runtime totals, the kind of stats MAL never gave me a way to compute.

That’s the easy version of the story. The actual migration was worse than I expected.

The completion-date problem

For ten years I marked anime “completed” without setting a finish date. I assumed MAL would record the date I made the change — that’s what any sane database would do. It doesn’t. If you don’t enter a date manually, the field stays empty.

Importing my MAL export into AniList preserved exactly that: thousands of completed entries with no finish date. Useless for any time-based analysis.

malexport and Selenium

Fortunately someone hit this problem before me. malexport walks through every anime and manga page on your MAL profile via Selenium, scrapes the activity history — which does record when you changed status — and extracts the implied completion timestamp.

It is exactly as slow and fragile as it sounds. It’s also the only way.

A few hours of headless Chrome later, I had a real per-entry timeline.

Backfilling AniList over GraphQL

The official MAL XML import to AniList had already created the entries, just without dates. So I wrote GraphQL mutations to walk every imported entry and update its completedAt field with the timestamps from malexport.

Once that landed, the data was finally what it should have been the whole time: every anime, with the date I actually finished it.

What I got

The migration is done. The analytics app works. And I’m no longer locked into a service that doesn’t let me touch my own data — which, ten years in, turned out to be the only thing I actually cared about.

← All posts