traversing the word graph, with links corresponding to characters that are shared between words (blue is when the reading is also the same, red is when it isn't)
all of this used to be implemented in "userspace" in about 1kloc of rust in a previous version—turns out you can do it with less than 150 lines of sql, which also happens to be several times faster
swann retweeted
i made the gofundme for my legal expenses for my refugee case here, thank you gofundme.com/f/help-me-pay-l…
swann retweeted
I love—and keep encountering—stories of people from radically different cultures encountering modern/western cities for the first time and being unimpressed except for one specific thing.
Do you know of any other stories like this?
i read a short story by kawabata yesterday evening (titled 心中, if you want to look it up) that completely, como se dice, bodied me despite being less than two pages long
rather pleased with how the furigana annotator has turned out
it's not in the screenshot, but this also tracks what rule/heuristic was used to match the dictionary ("underlying") reading to the word (e.g. かた -> がたvia rendaku, オー -> おう is "LongVowelMark", and so on)
if i had a nickel for every time an otherwise decent book from the 90s has suddenly hit me with "altaic is real btw"
serde is excellent as long as you can express exactly what you like with some combination of attributes on/in the definition of your type (which is most of the time) but the moment you go beyond that and have to write an impl by hand the ergonomics of it seems to fall off a cliff
























