[datastore] Fix #3743

This change adds support for key == NULL to the datastore plugins
and replaces the offset argument with a next_uid and random arguments to
increase performance in the key == NULL case.

With the offset argument a datastore plugin would have to count all
matching keys before fetching the key at the right offset, which would
iterate over the entire database in the case of key == NULL.

The offset argument was used in two ways: to iterate over a set of
matching values and to start iteration at a random matching value. The new API
seperates these into two arguments: if random is true it will return a
random matching value, otherwise next_uid can be set to uid + 1 to return the
next matching value.

The random argument was not added to get_zero_anonymity. This function
is used to periodically insert zero anonymity values into the DHT. I
don't think it's necessary to randomize this.

1 files changed, 2 insertions, 2 deletions

diff --git a/src/datastore/test_plugin_datastore.c b/src/datastore/test_plugin_datastore.c
index 9b85d57da..94d93aac6 100644
--- a/src/datastore/test_plugin_datastore.c
+++ b/src/datastore/test_plugin_datastore.c
@@ -64,7 +64,6 @@ struct CpsRunContext
   enum RunPhase phase;
   unsigned int cnt;
   unsigned int i;
-  uint64_t offset;
 };
 
 
@@ -308,7 +307,8 @@ test (void *cls)
                 "Looking for %s\n",
                 GNUNET_h2s (&key));
     crc->api->get_key (crc->api->cls,
-                       crc->offset++,
+                       0,
+                       false,
                        &key,
                        NULL,
                        GNUNET_BLOCK_TYPE_ANY,