- added missing <html> and <body>
- fixed non-working CSS link, added CSS link to all pages
- replaced layout by <br> and with structural tags with CSS
- removed superfluous output, notably in BeanBrowseResource
- <code>-ify variable names in ScriptResource
- added special handling of NaN in doubleToString()
* SurtPrefixedDecideRule.java
make class implement Checkpointable, save surt prefixes to json on checkpoint and load them on recover
* profile-crawler-beans.cxml
make main SurtPrefixedDecideRule a top-level bean so that it can be checkpointed
* Checkpoint.java, CheckpointService.java
add some FINE level logging during checkpointing and recovery
Note that the name of the engine may be different than the name of the language
it parses. A good example is the Rhino script engine that parses javascript.
* PathSharingContext.java
new member variable ConcurrentHashMap data and accessor getData()
* ScriptedProcessor.java, ScriptedDecideRule.java
make appCtx available to scripts; also remove unused member sharedMap
* ActionDirectory.java
formatting fix
default false is old behavior, skip candidates-handling of outlinks from response codes <200 and >=400
if true, these outlinks will be treated the same as others (get scope-tested and enqueued)
* FetchDNS.java
recordDNS() - use local variable for dummy read buffer, instead of member variable that was shared among threads, since RecordingInputStream.read() used the passed buffer as a work area
(using only path)
(RobotsPolicyTest) added for case that failed pre-fix
(RobotsTxtTest) added case similar to problem (but was already OK at
RobotsTxt level)
(RobotsPolicy, *RobotsPolicy) rename method to reflect new behavior
BaseQueuePrecedencePolicy.basePrecedence) doesn't stick
(CrawlURI) don't null overlayNames in processingCleanup
(WorkQueueFrontier) move queue precedence recalc to only in handleQueue,
not all (post-wake) reenqueues
* HostsReport.java
write() - remove obsolete code described by this no longer correct comment - "StatisticsTracker doesn't know of zero-completion hosts; so supplement report with those entries from host cache" - StatisticsTracker does know of zero-completion hosts
* FetchWhois.java
fetch() - record server ip address in CrawlURI attributes list
* CoreAttributeConstants.java
new string constant A_WHOIS_SERVER_IP, key for whois server ip address CrawlURI attribute
* WARCWriterProcessor.java
writeWhoisRecords() - include warc header WARC-IP-Address
* SheetOverlaysManager.java
getOverlayMap(String) - return null if sheet missing instead of triggering npe
* KeyedProperties.java
get(String) - check for null return value from getOverlayMap() and log warning
when starting in background, make better effort to write pid to $HERITRIX_HOME/heritrix.pid on successful start, plus this change:
- echo -n "`date` Starting heritrix"
+ echo -n "`date` Heritrix starting (pid $HERITRIX_PID)"
* KryoBinding.java
wrap ObjectBuffers with WeakReference so that can be garbage collected if necessary (when recreated they'll be back at the default size of 16k)
Reported by Kenji who says, "My guess is that root cause is Recorder.calcRecommendedCharBufferSize():
return Math.min(inStream.getRecordedBufferLength()/2,(int)inStream.getSize());
URL above returns HTML larger than 5GB (infinite smileys!! what the heck), and (int)intStream.getSize() became negative."
* Recorder.java
- return Math.min(inStream.getRecordedBufferLength()/2,(int)inStream.getSize());
+ return (int) Math.min(inStream.getRecordedBufferLength()/2, inStream.getSize());
Travis Wellman