You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@jena.apache.org by bu...@apache.org on 2018/06/04 15:44:47 UTC

svn commit: r1030768 - in /websites/staging/jena/trunk/content: ./ documentation/tdb2/tdb2_cmds.html

Author: buildbot
Date: Mon Jun  4 15:44:46 2018
New Revision: 1030768

Log:
Staging update by buildbot for jena

Modified:
    websites/staging/jena/trunk/content/   (props changed)
    websites/staging/jena/trunk/content/documentation/tdb2/tdb2_cmds.html

Propchange: websites/staging/jena/trunk/content/
------------------------------------------------------------------------------
--- cms:source-revision (original)
+++ cms:source-revision Mon Jun  4 15:44:46 2018
@@ -1 +1 @@
-1832858
+1832860

Modified: websites/staging/jena/trunk/content/documentation/tdb2/tdb2_cmds.html
==============================================================================
--- websites/staging/jena/trunk/content/documentation/tdb2/tdb2_cmds.html (original)
+++ websites/staging/jena/trunk/content/documentation/tdb2/tdb2_cmds.html Mon Jun  4 15:44:46 2018
@@ -182,22 +182,25 @@ h2:hover > .headerlink, h3:hover > .head
 
 
 <p>All TDB2 loaders can update datasets and do not have to work on an empty
-dataset.  However, all but the basic loader work by manipulting the low
-level datatstructures, and are tuned to large changes of data. They do
-not provide perfect transaction isolation in case a load goes wrong for
-some reason. The multiphase loading operations use partial transactions
-which can leave the database in a strange state.</p>
+dataset.  However, only the basic and sequential loader are fully
+transactional in the presense of crashes. The other loadrs, while
+faster, work by manipulting the low-level datastructures, and are tuned
+for large changes of data. They do not provide perfect transaction
+isolation in case a load goes wrong for some reason. The multiphase
+loading operations use partial transactions which can leave the database
+in a strange state.</p>
 <p>When working with large data to load, it is advisable to check it
-complete first with <code>riot --validate</code>. Parse errors during loading can
+completely first with <code>riot --validate</code>. Parse errors during loading can
 lead to inconsistent indexing. Fixing bad data, even if legal RDF, such
 as bad lexicial forms of literals or bad URIs, is much easier before the
 data is in the database.</p>
-<p>Because loading in very hardware dependent, the right choice for any
+<p>Because loading in hardware dependent, the right choice for any
 situation can only be found by trying each loader to see what works best
-and the notes below are just initial guidance.  The default choice is
+and the notes below are only initial guidance. The default choice is
 a reasonable starting point. Closing all applications to release their
 memory and not use CPU improves the loading process performance.</p>
-<p>Loader options:</p>
+<h3 id="loader-options">Loader options<a class="headerlink" href="#loader-options" title="Permanent link">&para;</a></h3>
+<p>The choice of loader is given by the optional <code>--loader</code> argument.</p>
 <p><code>--loader=basic</code></p>
 <p>The basic loader loads data as a single transaction into the dataset on
 a single thread. It is suitable for small data and also for