Yothalot-PHP 3.1.0

• Fixed compatibility with yothalot library, splitsize can now longer be set (it now always is 10mb)
• Refactored tempqueue to make it possible to add other feedback channels in the future (which we need for tcp based feedback)
• The pool now works with the more abstract concept of handlers instead of rabbitmq connections, to allow the future additional of alternative feedback channels
• Added first setup for the listener class, that allows feedback to be sent over tcp/ip instead of rabbitmq
• When the tcp parameter is provided, it is now possible to wait for an answer via tcp on a random port
• Removed recompress feature from the output::flush() method because it is also removed from the underlying c++ library

Yothalot-PHP 3.0.0

• prevent exceptions from bubbling up in case a directory does not exist
• fixed extension to be compatible with updated yothalot library (but added a dependency on unavailable private copernica libraries)
• updated default config file
• allow datasize shortcuts (like 1MB) in ini file
• the yothalot address can now be supplied in amqp:// format in the config file and as parameter to the job instead of supplying all parts individually
• fixed compile errors
• fixed wordcount example
• work in progress on nosql support for the yothalot php extension
• fixed a couple of todos for directory class
• work in progress on using nosql for file storage
• removed weird characters
• fixed allocation issue
• we now pass the nosql connection to the writer task, and the job::directory() method now always returns an directory that exists
• the yothalot.maxcache ini setting is now parsed as human-readable data size (like "1MB")
• fixed running finalization algorithm in client
• the input files for writer tasks that are executed by the client are now automatically removed
• no longer use readdir_r because it is going to be deprecated
• the nodes now use the same cache settings that were used on the client that started the mapreduce process
• renamed core class to rabbit class, added cache object to the connection class
• fixed serialization for connection object
• the cache settings that are passed to the yothalot connection object are now passed to the actual jobs, they no longer fetch settings from ini file
• the local running finalizer now uses the cache settings from the connection object
• cache object no longer holds a target object, because we do not need in in all places
• the rabbit data is no longer serialized when a job is serialized
• removed couple of todos
• cache settings are now passed to the yothalot server
• better error message
• internal cancelled property in the tempqueue should be set earlier to prevent double cancels
• for local finalizing, we need to change to the appropriate output dir
• removed the TupleHelper class, and refactored it to seperate classes
• added the RecordReduce class that allows to process records from the logfile directly

Yothalot-PHP 2.8.0

• First implementation of the Yothalot\Pool class, that allows to wait for the result of multiple jobs at the same time
• Tempdir can now be set using the yothalot.temp-directory ini file setting (default = /tmp)
• Added Descriptors class that encapsulates all descriptors used by a connection
• Split loop and descriptors, to allow future support of multiple amqp connections
• Allow callbacks to be passed to the event loop
• json::object is internally already implemented as a sort of shared-ptr, no need to wrap it into another shared-ptr
• Refactored stuff to make it easier to run multiple jobs in parallel, changed the return value for job::start()
• Added method to fetch a job from a queue without waiting, added method to get pool size
• Allow running the finalizer in the client
• Tempqueue channel is now closed after result has been read, to fix the connection::flush() method that does not expect channels to survive
• When a job is serialized/unserialized, it still is possible to run the finalizer locally
• The _cancelled member was not initialized on construction

Yothalot-PHP 2.7.1

• Fix issue with getting the correct algorithm for serialized objects
• Fix unserialize of job

Yothalot-PHP 2.7.0

• Added support for limiting record count in mapper tasks
• Add option to set the split size for mapper input files and check whether the the maximum number of bytes is a multiple of this split size
• Reworked error objects to extend from the regular result objects
• Pass exception from missing mount point on to PHP userspace
• Added Input\seek() method
• Updated extension to new PHP-CPP api
• Fixed the makefile for newer PHP versions
• Removed dependency on React CPP

Yothalot-PHP 2.6.0

• Fix C++ exceptions that bubbled up to PHP space if a Yothalot\Output object was not correctly constructed
• Use the pid of the process as a prefix to avoid getting the same 'unique' id, resulting in failure of directory creation
• Update extension to be compatible with updated yothalot library (input object no longer supports start and size parameter, but a strict parameter)
• Add dependencies to makefile
• Add the recompress option to the Yothalot\Output::flush method

Yothalot-PHP 2.5.0

• We no longer have a version 1 and 2 of mapreduce, there's a single MapTask now.
• Added backwards compatilibity for the kvmapper
• Added a Yothalot\Result interface for us to implement
• A Yothalot\Input doesn't have to be a log file.. just a yothalot file
• Fix issue with add and race jobs

Yothalot-PHP 2.4.0

• fix issue with flushing an output object
• remove unused task property
• add a member to input to check if the input file is valid
• add next method to input class
• make version 1 default if version is not set

Yothalot-PHP 2.3.1

• Take into account that we could have an error object in our result json

Yothalot-PHP 2.3.0

• Removed calls to Php::Namespace class (it appears to be broken)
• Implemented an ErrorResult object
• Don't emit empty warnings because that doesn't make much sense