[Wikisource-l] Wikistats seens to be inaccurate for Wikisources

[Erik Zachte]

Follow up: I see nl wikisource was not a good example.

In fr the namespaces Page and Livre are indeed defined, unlike at nl.
Hence these namespaces are counted separately.

     <case>first-letter</case>
       <namespaces>
       <namespace key="-2">Média</namespace>
       <namespace key="-1">Spécial</namespace>
       <namespace key="0" />
       <namespace key="1">Discuter</namespace>
       <namespace key="2">Utilisateur</namespace>
       <namespace key="3">Discussion Utilisateur</namespace>
       <namespace key="4">Wikisource</namespace>
       <namespace key="5">Discussion Wikisource</namespace>
       <namespace key="6">Image</namespace>
       <namespace key="7">Discussion Image</namespace>
       <namespace key="8">MediaWiki</namespace>
       <namespace key="9">Discussion MediaWiki</namespace>
       <namespace key="10">Modèle</namespace>
       <namespace key="11">Discussion Modèle</namespace>
       <namespace key="12">Aide</namespace>
       <namespace key="13">Discussion Aide</namespace>
       <namespace key="14">Catégorie</namespace>
       <namespace key="15">Discussion Catégorie</namespace>
       <namespace key="100">Transwiki</namespace>
       <namespace key="101">Discussion Transwiki</namespace>
       <namespace key="104">Page</namespace>
       <namespace key="105">Discussion Page</namespace>
       <namespace key="112">Livre</namespace>
       <namespace key="113">Discussion Livre</namespace>
     </namespaces>
   </siteinfo>

In http://stats.wikimedia.org/wikisource/EN/TablesWikipediaFR.htm
in the first table you will see counts for namespace 0 only.

In the table 'Database records per namespace' on the same page you can 
see counts for 104 separately. (112 is missing).

Do you want the total of all three namespaces 0+104+112) in column E in 
the first table? I can treat 104 and 112 as namespace 0 then. It will 
influence counts in all columns.

Treating them as one in the first table and separately in 'Database 
records per namespace' is doable. They will show up in
'Database records per namespace' in separate columns then.

-----------

If defined will the numbers for the keys always be 104 and 112? In that 
case I'd rather not harvest the codes from a html page or through the 
API, if they are directly available in the dump.

Cheers Erik Zachte

Luiz Augusto wrote:
> Hi Erik, I'm crossposting this message to the wikisource-l, if anyone is 
> interested to give some inputs.
> 
> The http://stats.wikimedia.org/wikisource/EN/TablesDatabaseWords.htm 
> seens to be inaccurate. Apparently your tool compute only words in the 
> main namespace. It may works for projects like Wikipedia and theirs very 
> long talk pages at the namespace Project: on some subjects (such as 
> deletion requests). But it doens't work for Wikisource for two main reasons:
> 
> 1) Some subdomains have custom namespaces for short biographies and list 
> of works by author (en, it, pt and others), some have it on the main 
> namespace (fr, de, es and others). This is a minor issue, since the 
> amount of words on those pages is small
> 
> 2) Some Wikisources (de, fr and en, according to 
> http://wikisource.org/wiki/Wikisource:ProofreadPage_Statistics ) have 
> large amount of contents in a custom namespace devoted to the 
> ProofreadPage Extension ( 
> http://www.mediawiki.org/wiki/Extension:Proofread_Page ). This content 
> is displayed on main namespace within page transclusion (see 
> http://en.wikisource.org/w/index.php?title=35_Sonnets&action=edit 
> <http://en.wikisource.org/w/index.php?title=35_Sonnets&action=edit> for 
> an example).
> 
> Is possible to include the custom namespaces for all Wikisources on your 
> automated calculation tool?
> 
> [[:m:User:555]]

-- 
Erik Zachte
Data Analyst
Wikimedia Foundation, Inc.
E-Mail: ezachte at wikimedia.org