Ok, that means that I can flip the paywall on then.
Then I’ll start removing features “to make the database run smoothly”.
I learned from the best.
10 Likes
I pushed some tweaks to the search algorithm:
- In the kradfile source 食 and 良 are treated as completely separate radicals, even though 食 is obviously drawn like 𠆢+良. I added a “secondary match” so that 良 will also match 食. So now you can find 養う by looking for
- While I was at it I also made it so 良 and 艮 would match each other since it can be easy to mix them up in some fonts. So now you can find 即刻 by searching for
- Vocab starting with お・ご・御 is annoying because some of those words have a dedicated vocab entry (ご覧, お手洗い) while in other situations it can just be a polite marker for a random noun that does not usually come with it (like I just encountered お戻り in a game, which doesn’t have a dedicated dictionary entry, it’s just 御 + 戻り). But if you don’t know the noun in the first place, how can you guess?
So I made it that if your search starts with お・ご・御, the search will consider it optional:
Note that, as always, exact matches should always be given priority, so hopefully it shouldn’t degrade the quality of the results too much with irrelevant matches.
Please let me know if you think that it causes weird matches in some cases.
6 Likes
I DIDN’T DO ANYTHING
Oh actually I reduced some margins I think, so it may have been enough to fit one more kanji on the line.
But I didn’t change the order so technically I didn’t break my promise. No refunds.
4 Likes
/ragequit
4 Likes
I pushed a small change:
- I added the Kun/On readings to the kanji in the vocabulary info:
- From now on every day at midnight one random radical will change place.
5 Likes
Nooooooooooo.
3 Likes
えっ、その「葡萄」の読み方は見たことがなかったよね。
4 Likes
I guess I’ll have to go back to trying to make it run locally so that I’m not affected by this last new feature 
4 Likes
I didn’t even screenshot the right word, I wanted 薔薇色 which I encountered in Trails in the Sky this morning because I thought it was cool that I could find it with:
Except of course that I also cut the query string from the screenshot so it’s now completely random…
According to the wiktionary えび is an obsolete reading for 葡萄.
4 Likes
ばら I can read it from long time ago, but still can’t write it 
I should definitely add it to my exercises list．
I can’t easily relate the components to kanji I know, unlike 蜘蛛、蟹、鰐 or even 竃 (虫知朱角刀牛咢穴土縄(の旁) are all simple blocks I know, but for ばら…
EDIT: 薔薇 are not even in ringotan… but thanks to this thread (and so, a bit indirectly also thanks to kankan) I will tackle it: 艹, a weird 坐, and 回．For the second one, it will be easier once I also learn to write “microbe”… or, I can now write 微生物 thanks to the rose: 艹彳山 almost 冗(but 一几 actually) and that old 攵 buddy
EDIT2: it seems the thing below the mountain can be written 一ル or 一几. Actually the number of strokes is the same, but visually it seems to be one more with 一几. Wikipedia says that version 一几 is simplified Chinese, yet that is what is used in Japanese fonts for 薇 while 微 uses the traditional Chinese version with ール. Well, as my 几’s really suck, I will opt for the easier to draw 兀.
4 Likes
It’s insane how “unstandardized” kanjis are as you dig deeper and deeper outside of jouyou.
d I added 喰 to my kanji drawing anki deck yesterday and at first I expected that it would just be 口 + 食 but if you look closely stroke 6 is actually drawn horizontally which I believe is an older variant of 食 that’s maintained in this particular kanji:
I mean even within jouyou it’s a bit of a mess frankly. So you have 戔 simplified into 㦮 in kanji like 残, 桟 or 践 but not in 箋. You have to use the so-called, nonstandard “extended shinjitai” to get 䇳.
Ditto for 龍, simplified into 竜, but 籠 and 襲 keep the old, much more complicated component while 篭 is again extended shinjitai and the simplified form of 襲 doesn’t seem to exist at all?
There are so many of those. 暇 still uses 叚 while 假 got simplified to 仮. Why not simplify every occurrence? Is it really a simplification if it’s partial and effectively makes things trickier? 假 actually makes more sense if you look at the onyomi…
3 Likes
For 食言令 etc, it’s not a big deal, that top dot can be written both ways (handwriting tend to use the vertical variant).
It is more annoying with things like 飲飴, or 辶, all major fonts I see have inconsistent style. It’s like having a Latin font in sans serif, but with j,v and w serifed, because the education ministry didn’t included a picture of them in some table.
(Thanks to your message, reminded me that my first 五十音 chart had 平假名 as title)
Japanese shinjitai is more based on visual rendering than preserving lexical consistency. Several ambiguities have been introduced, like 芸(in Chinese they made 芸 and 艺 disrincts, so the meanings (and sound) of the original 芸 (a plant name) and simplification of 藝 don’t get mixed.
PS: akebi app has various stroke order schemas, for example
4 Likes
I have a question about the ordering of the results. When I search for the 唱 kanji:
Then 唱える is quite far down in the results even though it is an exact match (kanji-wise) and Top 4k according to the frequency data. Why isn’t it higher up in the ordering?
1 Like
Because it has two extra characters not in the query, so something like 暗唱 with only one extra char is considered a better match.
The relative frequency does favor the more common 唱える but the extra character penalty more than overrides it. I also don’t distinguish kana and kanji for this scoring.
All of this is completely arbitrary of course.
2 Likes
Ah, okay, now I understand. Yeah, you have to choose some kind of ordering and this makes sense as well.
Have I told you how awesome this is? This is awesome. You should feel good. yay 
4 Likes
Happy that it’s useful!
3 Likes
Btw, I’ve been using this since @Akashelia mentioned it recently and this is an awesome tool for manual lookups. No more radical search on jisho \o/
4 Likes
I just pushed a bugfix because I got frustrated while reading bocchi: 「示」い wouldn’t find 祓い!
It turns out that 示 in 祓 is actually just a variant of ⺭ and using that as component in the search did return the kanji:
So I thought that I just had to add an exception to also search for ⺭ when the query contains 示.
Except I had already done that!
So what was the issue then?
"礻" == "⺭" // -> false
The first character is “CJK Unified Ideograph-793B”, the second is “CJK Radical Spirit Two”.
The 2nd character is the one I use everywhere in my decomposition, but for some reason I managed to copy/paste the first character in my table of secondary components for 示.
Now it’s fixed:
Unicode is fun.
8 Likes