On 2012-01-20 17:16, Bart van der Schans wrote:
On Fri, Jan 20, 2012 at 3:44 PM, Bart van der Schans
<[email protected]> wrote:
On Fri, Jan 20, 2012 at 3:39 PM, Julian Reschke<[email protected]> wrote:
On 2012-01-20 15:22, Bart van der Schans wrote:
Hi all,
I was wondering about how the ConsistencyCheckerImpl fetches all the
node ids. The first set of node ids is fetched with
"pm.getAllNodeIds(null, NODESATONCE)" in the internalCheckConsistency
method. All subsequent sets are fetched with "pm.getAllNodeIds(lastId,
NODESATONCE)".
In the BundleDbPersistenceManager the method
getAllNodeIds(NodeId,maxCount) uses the "bundleSelectAllIdsSQL" if the
first parameter is null otherwise the "bundleSelectAllIdsFromSQL"
query. The "bundleSelectAllIdsSQL" query doesn't use ordering and the
"bundleSelectAllIdsFromSQL" does.
Now my question is the following: souldn't the "bundleSelectAllIdsSQL"
also use an order clause to make sure the correct batch of node ids is
fetched the first time? Or is that somehow implicitly guaranteed?
Good question. I was confused about that as well. I tried and stepped
through he code and everything seem to worked as advertised, so I stopped
worrying about it :-)
I think (at least on mysql) we're just lucky that the natural ordering
is by NODE_ID:
mysql> SELECT hex(NODE_ID) FROM VERSION_BUNDLE ORDER BY NODE_ID LIMIT 0,10;
+----------------------------------+
| hex(NODE_ID) |
+----------------------------------+
| 000005E87E6D402BA27CDF41E7999D8A |
| 000007DC02CF48BA8FFEDA5F566A3A98 |
| 00000C4A81744934B4F0CB55C9967378 |
| 000016C9E4464C4FA7B6F4B9CF2F9903 |
| 00001D5A1C6F4339B317BB436AA54431 |
| 0000257D67AB4F1EA4DDEDB84BB13039 |
| 00002C51F22145B89F59E882BB505036 |
| 00002F2AFC2E4842863C360B68A36BDA |
| 0000354F151945CEA953D62C922A4FAC |
| 00003C0254C8471E8349C0B6F0DEFDC2 |
+----------------------------------+
10 rows in set (0.00 sec)
mysql> SELECT hex(NODE_ID) FROM VERSION_BUNDLE LIMIT 0,10;
+----------------------------------+
| hex(NODE_ID) |
+----------------------------------+
| 000005E87E6D402BA27CDF41E7999D8A |
| 000007DC02CF48BA8FFEDA5F566A3A98 |
| 00000C4A81744934B4F0CB55C9967378 |
| 000016C9E4464C4FA7B6F4B9CF2F9903 |
| 00001D5A1C6F4339B317BB436AA54431 |
| 0000257D67AB4F1EA4DDEDB84BB13039 |
| 00002C51F22145B89F59E882BB505036 |
| 00002F2AFC2E4842863C360B68A36BDA |
| 0000354F151945CEA953D62C922A4FAC |
| 00003C0254C8471E8349C0B6F0DEFDC2 |
+----------------------------------+
10 rows in set (0.00 sec)
But that doesn't have to be the case with other databases. Maybe
somebody can try similar queries on another database?
Does anybody have an objection to adding an order clause to the
"bundleSelectAllIdsSQL" statement? If not, I'll create an issue and
add it.
Regards,
Bart
Sounds good to me.
Best regards, Julian