Re: [WikiTrust] problem installing wikitrust

[Luca de Alfaro]

Hi Hoda!

On Tue, Aug 4, 2009 at 4:57 PM, Hoda Sepehri Raad <sepeh...@gmail.com> wrote:

Thanks, you were right. That obvious solution which I had not try, made it work.

Now, first I was wondering if the script for importing the data (load_data.py) can import the data to a non-empty database or not. Because using the "mwdumper" file alone, I could import the data successfully to an empty database, but the documentation of "mwdumper" says that some tables such as "text", "revisions", etc. should be empty before importing.

load_data.py can clear the db tables before calling mwdumper, so it should solve this problem.
 

Second, I got a problem working with "load_data.py". The file db_access_data.ini which should contains the db information asks for the name of the database. I filled it with the correct value, yet it complains that: "ERROR 1146 (42S02) at line 8: Table 'wiki_analysis.text' doesn't exist". The problem is that my tables in this database have a prefix. So, the name of for example "text" table is "wp_phil_text". I changed the database name to "db_name + prefix", but it complained that the database name is wrong. So, what should I do for solving this?

It may well be the case that load_data.py does not handle extensions properly.  It might also be that you specify only part of the extension.  It is purely a test script we use to set up debugging instances of wikis, so it's not really part of the "polished" code.
You can file a bug report (at google code for wikitrust, or look at wikitrust.sourceforge.net and follow the links).
You have then various options:

• You can fix the bug yourself (it is fairly simple python code), send us the patch / diff, and we will be happy to include it and give you a big Thanks.  Look at the code.  Replace the curs.execute with print statements, and see what you get.  Does it look like the correct sql to empty the tables?
  
• You can just look at load_data.py, at how it truncates the tables, and do it by hand.
  
• You can wait a bit till we fix the issue; for us it is low priority as it is not part of what we consider to be the "supported" or "polished" part of WikiTrust.
  

Third, I got a little bit confused. If I want to use the "author reputation" analysis, should I first import the data to the database using any of the previous methods, or simply giving the xml file to the analyzer step will be enough?

You can do either.  If you import the data into the database, you can then launch ./eval_online_wiki as explained in the README file, and you are left with a wikitrust_user table that contains everybody's reputation.  You also get a lot of other data: for example, you can discover due to what revisions every author gained / lost how much reputation.

There is also a faster way to get this information from the xml dump of a wiki, but it involves several steps; you can read about it in the batch/README directory. 
 

I tried with giving the sample xml file in test-data folder to "computing the reduced statistics" step, and I got two files: "*.stats", and "*.stats.info.xml" , but I did not get any "*.out" file as described in README file. I should add that I got "git not found" in the middle of producing these two files. Is this O.K?  The ".stats" file is the same as the ".stats" file in "test_input" directory under "batch" directory" when trying with one of the xml files provided there.

Thanks!  The extension changed from .out to .stats, and I will fix the README file.
I am aware of the "git not found" error.
It does no harm, and I fixed it already in the development branch.

Btw, all the code is on sourceforge now (we are considering also github, but for the moment, it is on sourceforge).  There you can download, using git, the most recent version of WikiTrust, if you wish.  See http://wikitrust.sourceforge.net .

Best,

Luca
 

Thank you again very much for your time and  help,
--Hoda

On Wed, Jul 29, 2009 at 7:41 PM, Luca de Alfaro <lu...@soe.ucsc.edu> wrote:

It may seem obvious, but did you try

Make clean ; make all ; make allopt

?

Luca
---
Sent from mobile

On Jul 29, 2009 5:58 PM, "Hoda Sepehri Raad" <sepeh...@gmail.com> wrote:

Thank you very much for your help. Using godi was a great way. Just could not find "ocaml-pcre" in godi list of packages. After installing all the others, I tried making the "analysis" directory under "batch" directory(I want to use batch mode), I got a new error:

File "textbuf.ml", line 1, characters 0-1:
Could not find the .cmi file for interface textbuf.mli.

No idea what this error could be related to.

Thanks again for your help.
--Hoda

On Wed, Jul 29, 2009 at 9:30 AM, Luca de Alfaro <lu...@soe.ucsc.edu> wrote: > > I think the package...