0) Problem

The elwikisource-20150602-stub-meta-current.xml.gz violates UNIQUE KEY constraint.
This makes trouble for INSERT commands.

1. Database schema

CREATE TABLE page (
page_id int(10) unsigned NOT NULL AUTO_INCREMENT,
page_namespace int(11) NOT NULL,
page_title varbinary(255) NOT NULL,
...
PRIMARY KEY (page_id),
UNIQUE KEY name_title (page_namespace,page_title),
KEY page_random (page_random),
KEY page_len (page_len),
KEY page_redirect_namespace_len (page_is_redirect,page_namespace,page_len),
KEY page_latest (page_latest)
) ENGINE=InnoDB CHARSET=binary;

2. Dump contents

(shell) emacs elwikisource-20150602-stub-meta-current.xml.gz
...
<namespace key="108" case="first-letter">Συγγραφέας</namespace>
...
<page>
<title>Συγγραφέας:Πλάτων</title>
<ns>0</ns>
<id>1422</id>
<redirect title="Πλάτων" />
...
<page>
<title>Πλάτων</title>
<ns>0</ns>
<id>21023</id>
<redirect title="Συγγραφέας:Πλάτων" />
...

3. Database table contents

If I drop the UNIQUE KEY constraint, then I can INSERT both pages:

mysql> select page_id,page_namespace,page_title from elwikisource.page where page_namespace=0 and page_title='Πλάτων';

But then we discover that Plato is not the only culprit. This dump file has ten such cases:

mysql> select page_namespace,page_title,count(page_id) as n from elwikisource.page group by page_namespace,page_title having n>1;

Or better yet

mysql> select a.page_id,b.page_id,a.page_namespace,a.page_title from elwikisource.page as a, elwikisource.page as b where a.page_namespace=b.page_namespace and a.page_title=b.page_title and a.page_id<b.page_id order by a.page_id;

4. Analysis

In nine cases, we see:

• a pair of <redirect>s that point to each other;
• where one <page> has the wrong value for <ns> (Συγγραφέας should be 108 not 0).

The tenth case, Main_Page, I can fix myself.

• When you set up MediaWiki, a default Main_Page is installed.
• When you INSERT an XML dump, a second Main_Page is installed.

The fix is to DELETE the first Main_Page. This I will do in the next version of WP-MIRROR.