Correctly read phrase frequencies for sorting

[yan12125]

Closes #334

[yan12125]

Well, no CI tests are run as travis-ci.org is ceased. With this change, all tests are still green on my machine.

[ShikiSuen]

@yan12125 What about GitHub continuous integration?

[ShikiSuen]

不是太懂 C，不知道這裡是否可以用 GetUint32。

[yan12125]

如果要用uint32，詞庫的格式以及libchewing要跟著改，用四個uint8_t表示頻率。目前我是以相容現有的libchewing為目標。

[yan12125]

What about GitHub continuous integration?

That's a good idea. If libchewing core developers are fine with GitHub Actions, I can try it in another PR.

[yan12125]

Rebased to test the change through GitHub Actions, which was added recently in #349.

[yan12125]

@kanru Mind to have another look?

Per #334 (comment), it was expected that init_database.c would be removed in favor of init_database.rs. It turns out that the latter uses a different file format - 32 bit integers for phrase frequency [1] rather than 24-bit integers as in init_database.c. As long as WITH_RUST=OFF is supported, init_database.c should be maintained. Thus, I think it's time to revisit this fix.

[1]

libchewing/tools/src/bin/init_database.rs

Line 92 in cebaeb3

let freq: u32 = match phrase.chars().count() {