I'm trying to filter out accented words if user searches for them in local database. But I have problems, namely with slavic letters ČŠŽ. In my SQLite database I have a field "title" with value: "Želodček"
If I try to select LOWER(title) I always get back the same value "Želodček" whilst other words are correctly lower cased. Only if the word begins with ČŽŠ then it doesn't get lower cased. This only persists with words which have leading accented letters.
Database records
Stomach
Želodček
Uppercase with UPPER()
STOMACH
ŽELODčEK
Lowercase with LOWER()
stomach
Želodček
I've already tried setting localization with setLocale() with no luck. I also tried different collation like NOCASE, UNICODE, LOCALIZED but nothing worked. I'm wondering why when lower cased the first letter is not lower cased and when upper cased other accented words are lowercase.
I've solved the problem with LIKE searches where I replace accented words with their lower cased counterpart. But I have problem with full text(FTS3) searching because I can't use the same trick with MATCH.
-- works but it's a hack
SELECT title FROM articles WHERE REPLACE(LOWER(title),'Ž','ž') LIKE '%želodček%'
-- can't seem to get it work
SELECT title FROM articles WHERE title MATCH 'želodček' COLLATE NOCASE
Is there any solution to this or is there a bigger problem?
Update: No optimal solution yet.
Un-optimal solution 1: I decided to deal with the problem directly by changing data in the select query. While this doesn't work for all cases (and I would have to cover all accents) it suits my case for now. So I'm posting it:
-- LIKE query
SELECT title FROM articles WHERE (REPLACE(REPLACE(REPLACE(LOWER(title),'Č','č'),'Š','š'),'Ž','ž') LIKE ? COLLATE NOCASE))
-- MATCH query (FTS)
-- In this case I programmatically replace searched word with 2 word variation (one that starts with lowercase and one that starts with uppercase) ie: title='želodček OR Želodček'
SELECT title FROM articles WHERE title MATCH ? COLLATE UNICODE
Un-optimal solution 2: As suggested by user CL. to insert in normalized form (didn't work for me because normalized form was basically the original unicode form). I took it futher and insert title stripped of of accents (basically ASCII form). This is maybe better than solution one in ways of general solution. Since I only cover some accents in the first. But there are downsides:
Here's the Java code for it:
// Gets you the ASCII version of unicode title which you insert into different column
String titleAsciiName = Normalizer.normalize(title, Normalizer.Form.NFD)
.replaceAll("[^\\p{ASCII}]", "");
LIKE never uses a custom collation .
FTS can use a custom tokenizer , but you have to check whether unicode61
is available in all Android versions you want to support.
The Android database API does not allow to create custom implementations of LIKE
or of a FTS tokenizer. You might want to store a normalized version of your strings in the database.
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.