23 Jan
2010
23 Jan
'10
12:17 p.m.
On Jan 23, 2010, at 12:08 PM, Sergiu Dumitriu wrote:
On 01/23/2010 09:01 AM, Vincent Massol wrote:
Using the full name is bad because it means a lot of parsing/serialization, that's why
we introduced entity reference. Previously all document names were manipulated as fullname
string so that's maybe why it was introduced.
On Jan 23, 2010, at 1:10 AM, Sergiu Dumitriu wrote:
I'm not sure who and why introduced it, but IMO it's normal. Previously
all the code that wanted to search using a document name had to either:
- manually split the fullname into space and name, with indexOf and
substring
- or use XWiki.getFullNameSQL(), which does a concat
The first one is boring to copy/paste all over the place, while the
second one hurts the performance. On 01/22/2010 07:38 PM, cjdelisle (SVN) wrote:
Sergiu can you explain why we store fullName in the DB since there's already page and
space. It's duplicated info which increases the risk of sync issues between the the
columns. In addition I don't really see any difference in retrieving one field or 2
fields (in term of string size the full name is actually longer than the 2 fields
separately so more time for comparison and serialization). Author: cjdelisle
Date: 2010-01-22 19:38:30 +0100 (Fri, 22 Jan 2010)
New Revision: 26312
Modified:
platform/core/trunk/xwiki-core/src/main/java/com/xpn/xwiki/store/XWikiHibernateStore.java
Log:
XWIKI-4754 parameterize query in getTranslationList
Modified:
platform/core/trunk/xwiki-core/src/main/java/com/xpn/xwiki/store/XWikiHibernateStore.java
===================================================================
---
platform/core/trunk/xwiki-core/src/main/java/com/xpn/xwiki/store/XWikiHibernateStore.java 2010-01-22
17:49:40 UTC (rev 26311)
+++
platform/core/trunk/xwiki-core/src/main/java/com/xpn/xwiki/store/XWikiHibernateStore.java 2010-01-22
18:38:30 UTC (rev 26312)
@@ -3024,11 +3024,12 @@
public List<String> getTranslationList(XWikiDocument doc, XWikiContext
context) throws XWikiException
{
- String hql =
- "select doc.language from XWikiDocument as doc where doc.space =
'" + Utils.SQLFilter(doc.getSpace())
- + "' and doc.name = '" +
Utils.SQLFilter(doc.getName())
- + "' and (doc.language<> '' or (doc.language is
not null and '' is null))";
- List<String> list = context.getWiki().search(hql, context);
+ String hql = "select doc.language from XWikiDocument as doc where doc.space
= ? and doc.name = ? "
+ + "and (doc.language<> '' or (doc.language
is not null and '' is null))";
doc.space and doc.name are explicitly used for historical reasons (in
the beginning doc.fullName was not stored in the database). For better
performance we could just use doc.fullName = ?. I don't know if you're right about the
performance of two fields versus
one longer, I believe the contrary is true. Databases don't work on
string comparison (when the proper indexes are in place), a search on
one index should be close to linear time, while joining two indexes
still requires a join, a costly operation.
Yes with indexes you're right but you can also have an index spanning several columns
(composite index).
All that said it's going to be hard to remove the fullname in the DB since it's
used directly in user's HQL queries. Unless there's a way to intercept these
queries easily after they are parsed by Hibernate and do some internal mapping.
-Vincent
Caleb, note
that if you wanted to be compliant with references (and not used the deprecated
doc.getFullName()) you'd need to use:
this.localEntityReferenceSerializer.serialize(doc.getDocumentReference())
Ah, yes, we'll all have to get used to the new APIs.
> For space only it's:
> doc.getDocumentReference().getLastSpaceReference().getName()
>
> For page it's:
> doc.getDocumentReference().getName()
>
> Thanks
> -Vincent
>
>>> + ArrayList<String> params = new ArrayList<String>();
>>> + params.add(doc.getSpace());
>>> + params.add(doc.getName());
>>> + List<String> list = search(hql, 0, 0, params, context);
>>> return (list == null) ? new ArrayList<String>() : list;
>>> }