- close stackfile inputstream at end of ChunkIterator
This should solve startup delay while unfinished crawl jobs exist (maybe also too many open file situation)
This organizes all urls to be loaded in separate queues for each host.
Each host separates the crawl depth into it's own queue. The primary
rule for urls taken from any queue is, that the crawl depth is minimal.
This produces a crawl depth which is identical to the clickdepth.
Furthermorem the crawl is able to create a much better balancing over
all hosts which is fair to all hosts that are in the queue.
This process will create a very large number of files for wide crawls in
the QUEUES folder: for each host a directory, for each crawl depth a
file inside the directory. A crawl with maxdepth = 4 will be able to
create 10.000s of files. To be able to use that many file readers, it
was necessary to implement a new index data structure which opens the
file only if an access is wanted (OnDemandOpenFileIndex). The usage of
such on-demand file reader shall prevent that the number of file
pointers is over the system limit, which is usually about 10.000 open
files. Some parts of YaCy had to be adopted to handle the crawl depth
number correctly. The logging and the IndexCreateQueues servlet had to
be adopted to show the crawl queues differently, because the host name
is attached to the port on the host to differentiate between http,
https, and ftp services.
instead of TreeMaps)
- enhanced memory footprint of database indexes (by introduction of
optimize calls)
- optimize calls shrink the amount of used memory for index sets if they
are not changed afterwards any more
The resource observer is now able to recognize free disk space AND
available space for YaCy. The amount of space which is assigned for YaCy
are defined in new settings in the configuration file.
Furthermore, there is now a cleanup process which deletes files in case
that an autodelete is activated. The autodelete is now BY DEFAULT ON if
the disk space is low, which means that YaCy starts to delete documents
when the disk is full!
- refactored all code which uses URIMetadataRow as standard for word
hash length and word hash ordering and moved that to the class 'Word',
becuase the class URIMetadataRow defined the old metadata data structure
and should be superfluous in the future
- removed unused methods from URIMetadataRow as preparation for further
removal of that class
where the size() and isEmpty() method is used only for statistics, which
happens at many locations in YaCy. If these methods are used for
structual reasons (like accessing the last element in an array) then it
may fail or cause other problems. As far as visible, this is not the
case.
- transformed log lines to String before they are stored because the
storage space is about 1:250 (45kb for one line before transformation,
180 bytes afterwards)
- this saves up to 10MB RAM so we can increase the number of lines to
1000 again.
the right content domain (i.e. identifying that it is an image, text
etc.) because it used the file extension and not an existing mime type
assignment.
- fixed the new setting that images shall be loaded for a better image
search.
- both fixes together makes it now possible to crawl
commons.wikimedia.org which makes use of 'funny' document names (i.e.
ending with .jpg while the document is html)
all unique links! This made it necessary, that a large portion of the
parser and link processing classes must be adopted to carry a different
type of link collection which carry a property attribute which are
attached to web anchors.
- introduction of a new URL class, AnchorURL
- the other url classes, DigestURI and MultiProtocolURI had been renamed
and refactored to fit into a new document package schema, document.id
- cleanup of net.yacy.cora.document package and refactoring
jdk-based logger tend to block
at java.util.logging.Logger.log(Logger.java:476) in concurrent
environments. This makes logging a main performance issue. To overcome
this problem, this is a add-on to jdk logging to put log entries on a
concurrent message queue and log the messages one by one using a
separate process.
- FTPClient uses the concurrent logging instead of the log4j logger
While the values for the reference evaluation are computed, also a
backlink-structure can be discovered and written to the index as well.
The host browser has been extended to show such backlinks to each
presented links. The host browser therefore can now show an information
where an document is linked. The new citation reference is computed as
likelyhood for a random click path with recursive usage of previously
computed likelyhood. This process is repeated until the likelyhood
converges to a specific number. This number is then normalized to a
ranking value CRn, 0<=CRn<=1. The value CRn can therefore be used to
rank popularity within intra-domain link structures.
id to be tested, but with a collection of ids. This will cause only a
single call to solr instead of many. The result is a much better
performace when testing the existence of many urls. The effect should
cause very much less IO during index transmission, both on sender and
receiver side.
process counter if an blocking thread dies. Added also a new column in
PerformanceConcurrency_p servlet to show the actual number of concurrent
processes.
- check geolocation coordinates and accept only those, which are
well-formed
- the solr push process does not stop crawling any more if after 20
requests to Solr Solr does not accept the record. Instead, a severe log
entry asks the user to create a bug request
adjusted to smaller and 1-core devices.
- the workflow processor now starts no process at all. these are started
as soon as parser/condenser/indexing queues are filled.
- better abstraction
references_external_i and references_exthosts_i. These can be used to
count and evaluate the number of external links to every web page. An
experimental ranking function can be i.e.:
div(add(references_internal_i,product(references_external_i,references_exthosts_i)),add(clickdepth_i,1))
- memory allocation problem in RowCollection: if memory is too low, do
not to try to increase by 1 because this leads to very long execution
time and at the end to the same OOM as if we allocate the memory at the
moment we need it even if the resource observer states that this memory
is not there. To compensate this, the increase size is reduced.
- an existing ranking servlet for solr was extended. It is now possible
to set boost values for fields, boost functions and boost queries.
- The ranking can have different instances, but currently only the first
one is used
- added an abstraction layer for fields which can be used for search and
those fields can be edited in the solr ranking configruation
- the ranking value from solr within the field score is used to combine
remote search requests, which all are created using the same locally
defined boost values
- reduced the number of fields which are used for search (makes it
faster)
- replaced some text fields by string fields (makes indexing faster)
- removed classes which had no use
- made a large number of experiments for a better ranking and created a
temporary setting which prefers hits inside titles
- adjusted also the RWI-based ranking computation to 'prefer title'
- made special cases like for portal search where no post-processing and
post-ranking is wanted: this keeps the original ranking order as done by
Solr
- fixed many bugs with old settings for ranking
- removed 'worker' processes
- no internal time-out behaviour: methods either are successful or
return null
- waiting is only done on top-level
- removed snippet-production; this is replaced by solr snippets
- removed statistics based on solr size queries (they had been VERY
long); the statistics (like suggestions or tag cloud) are now again
based on the old but very fast RWI index. In portal or intranet mode the
RWI index is usually switched off; if you like to have statistics again
then you must switch on the rwis again in this mode.
- fixed many bugs regarding correct page counter
The default schema uses only some of them and the resting search index
has now the following properties:
- webgraph size will have about 40 times as much entries as default
index
- the complete index size will increase and may be about the double size
of current amount
As testing showed, not much indexing performance is lost. The default
index will be smaller (moved fields out of it); thus searching
can be faster.
The new index will cause that some old parts in YaCy can be removed,
i.e. specialized webgraph data and the noload crawler. The new index
will make it possible to:
- search within link texts of linked but not indexed documents (about 20
times of document index in size!!)
- get a very detailed link graph
- enhance ranking using a complete link graph
To get the full access to the new index, the API to solr has now two
access points: one with attribute core=collection1 for the default
search index and core=webgraph to the new webgraph search index. This is
also avaiable for p2p operation but client access is not yet
implemented.
structure, but is not filled yet. To have the opportunity of a second
core, multi-core functionality had to be implemented to the
deep-embedded solr:
- migrated the solr_40 directory content to a subdirectory
'collection1'; the previously used default core is now called
collection1
- added solr_40/webgraph subdirectory as second core
- added a servlet configuration for the second core 'webgraph' in
/IndexSchema_p.html
- added instance handling as addition to solr connections: all solr
connectors are now instances of an solr 'instance' object; this required
a complete re-design of the solr embedding
- migrated also caching and sharding ontop of new instance handling
- migrated the search apis to handle now the access to a specific core,
the default core named 'collection1'
- migrated the remote solr search interface to access shards of cores;
for the yacy remote search the default core is now called 'solr'; using
the peer address as solr address
- migrated the solr backup and restore process: old backups cannot be
used after this migration!
- redesign of solr instance handling in all methods which access the
instances: they cannot hold copies of these instances any more; the must
retrieve the actuall connection object every time they want to write to
it (this solves also some bugs when switching the index/network)
- added another schema 'solr.webgraph.schema', the old solr.keys.list is
replaced by solr.collection.schema
