WXR importing duplicates categories

[takayukister]

In admin panel [Manage] - [Import], importing WordPress eXtended RSS (WXR) incorrectly generate duplicated categories with same cat_name.

For example, this occurs under condition below.

At exporting:

cat_ID	cat_name	category_nicename
3	書評	book-review

(書評 means book review in Japanese.)

And after importing this WXR file, categories will look like below.

cat_ID	cat_name	category_nicename
3	書評	book-review
4	書評	%e6%9b%b8%e8%a9%95

Two 書評 categories are generated. One has original "book-review" nicename and another has "%e6%9b%b8%e8%a9%95" which is sanitized string of "書評". And all posts in 書評 category belongs to 書評 category of "%e6%9b%b8%e8%a9%95".

It seems that this duplication occurs only when a category has category_nicename which is not equal to sanitized cat_name.

For easy reproducing of this behavior, I'll attach example WXR file. And I'll attach a patch for this.

[takayukister]

WXR file for example. UTF-8 encoded.

[rob1n]

Wouldn't it be easier to use category_exists()?

[takayukister]

​category_exists() depends on category_nicename to check existence of category. And it uses sanitize_title(cat_name) to get the category_nicename.

Under the example described above, sanitize_title("書評") returns "%e6%9b%b8%e8%a9%95" and because there's been no category with category_nicename "%e6%9b%b8%e8%a9%95" ever, another 書評 category (cat_ID:4) is created.

My patch uses cat_name instead of category_nicename to check category existence.

[rob1n]

(In [5246]) Use cat_name instead of cat_nicename when creating categories from import. Props takayukister. fixes #4040

[rob1n]

(In [5247]) Use get_var and only select cat_ID. see #4040

[takayukister]

I reopened to upload additional patch to correct two points below,

1. $post_cats should always include $cat_ID, regardless the category is existent or not.

2. Replace $post_ID to $post_id.

$post_ID is old ID on WXR file and $post_id is new ID. It's confusing naming (I'm sorry because I named $post_ID). It would be good to give them better names.

[rob1n]

(In [5252]) Some fixes for another fix. Props takayukister. fixes #4040