TODO.h

Go to the documentation of this file.

#ifndef TODO_H
#define TODO_H

//---------------------------------
//
//After dark or reference acquire, can't do another acquire
//
//Was deadlock problem - changed to recursive ReadWrite lock
//
//--------------------------------
//
//Enter QavrgAcquisition::initialize()loadReferenceData from /home/jennings/.qavrg/ref0
//QObject: Cannot create children for a parent that is in a different thread.
//(Parent is QavrgScriptingEngine(0x8caaf78), parent's thread is QavrgScriptingThread(0x8c3cf60), current thread is QThread(0x8af88a0)
//Leave QavrgAcquisition::initialize()QavrgSpecServer:: QavrgSpecServer::QavrgSpecServer
//
//Was problem with parent argument to QavrgServerThread
//
//--------------------------------
//
//Updating the graph frequently doesn't happen when it should
//Should work now...
//
//--------------------------------
//
//Want to remove 'preferences' dialog
//
//--------------------------------
//
//Display settings aren't saved
//
//--------------------------------
//
//Data don't scale correctly in simulated acquisition mode
//
//--------------------------------
//
//Fitting results don't update during scope mode acquisition
//QavrgResultsModel::resultsChanged() is emitted lots of times when acquisition ends.
//
// Problem was that waveforms were calculated by a QtConcurrent::run using a thread-safe
// random number generator.  The waveforms would only change when a new thread from
// the thread pool was used.
//
//--------------------------------
//
//Remove QavrgAcquisition::totalChannels()
//
//--------------------------------
//
//Resolve changes in QSpecServer
//Needs to be checked that it still works...
//
//--------------------------------
//
//Check that calculations are actually performed concurrently where possible - doesn't seem to be
//happening at the moment.
//The fitters read data from the acquisition object many times during execution, leading to extensive
//locking which stops the fitter threads from running in parallel.
//
//--------------------------------
//
//If you try to zoom the graph during scope mode, there is some kind of deadlock.
//(Only in real acquisition mode, it seems);
//
//--------------------------------
//
//Change QReadWriteLock to QMutex
//
//--------------------------------
//
//Change the fitted results table display:
//Display the peak height fits first
//Change the row number display to show the orbit and bunch number
//
//--------------------------------
//
//Change the main display bunch number to mean the filled bunch, rather than the bucket position.
//
//--------------------------------
//
//Calculate and display fitted data in main graph
//
//--------------------------------
//
//Adopt consistent channel numbering for human-visible purposes:
//
//--------------------------------
//
//Need to re-implement script "savedata" command, reduce number of columns
//
// Re-implemented, only save peak height
//
--------------------------------

Make THREAD_CHECK macro failure non-fatal, just print an error

--------------------------------

Provide some way of resetting the display and/or properties back to sane
default values.

--------------------------------

Change 'Result' terminology to 'FittingResult'

--------------------------------

Adopt consistent naming for set_ / get_ vs read / write and load / save
set/get for in-memory properties
read/write for files
load/save for file dialog user interface

//--------------------------------
//
//Perform acquisition accumulation into internal variables, copy results into
//visible variables and emit signals when completed.
//
//--------------------------------
//
//Add some zooming controls to the graphs
//
//--------------------------------
//
//Acquire Dark and Reference to put results directly into appropriate buffer without affecting
//raw data buffer.
//
//--------------------------------
//
//Added RCS ident strings to all files, and moved RCS log comments to bottom of files
//
//--------------------------------
//
//If the graph takes too long to update in scope mode the application will get stuck...
//Added a counter to keep track of outstanding updates and skip all except the last.
//Changed the acquisition wait loops to wait in chunks of 0.1 sec, to allow more
//responsive cancellation.
//
--------------------------------

Need to check that the actual acquisition behaves in much the same way as the simulated

--------------------------------

Need to check that the remote control interface works as expected

--------------------------------

Need to check file output routines

//--------------------------------
//
//Implement 'status' scripting function.  With no arguments returns instantaneous
//acquisition status, with an argument waits for up to that time for acquisition
//to complete then returns acquisition status
//
//--------------------------------
//
//Check 'isStored' when loading back property values
//Consolidate 'loadSettings' and 'saveSettings' routines in one place (static QavrgProperty ?)
//Allow settings to loaded and saved to user-chosen files as well
//
//--------------------------------
//
//Check fill pattern handling
//
--------------------------------

Improve error checking and handling in scripting interface.
//Make sure acquisitions don't get queued up.

//--------------------------------
//
//First acquisition run when number of samples and channels reduced outputs wrong numbers of
//columns.
//
--------------------------------

Initialization of fitter object depends on acquisition parameters which have not been
read from settings at the time the fitter object is constructed.

//--------------------------------
//
//Add user interface for setting number of samples and number of channels, including an easy
//way to reset them to default values.
//
--------------------------------

Handle acquisition timeouts better.

//--------------------------------
//
//Acquisition status should be changed when the acquireXXX functions are called, not when the event loop
//executes.
//

//--------------------------------
//
//Allow script execution to be interrupted
//
//--------------------------------
//
//Can't cancel acquisition any more. Data acquisition also never terminates, though dark and reference
//acquisition will finish eventually.
//
//Actually a size problem if nsamples was changed between dark/reference acquisition and data acquisition.
//Fitting routines crash in this case.
//Fixed.
//
//--------------------------------
//
//Acquisition repetitions after first are not handled properly
//Was bracket nesting problem in QavrgAcquisition::startAcquisition where the acquisition
//starting code was in the same block as the accumulator initialization code
//
--------------------------------

Consider moving data storage for each channel into the fitter object.

--------------------------------

Remote control interface still converts array return values into strings
i.e. remote_eval(HOST, "Array(2,3,4)")
returns the string "2,3,4" to spec

//--------------------------------
//
//Move QavrgProperty into qceplib
//
//--------------------------------
//
//Evaluate whether the simple expedient of using Qt::DirectConnection links between widgets
//and properties will work on all supported platforms.  It appears to be a simple way
//of avoiding the multiple queued updates problem that we are experiencing at present.
//Doesn't work, unfortunately.
// Have eventually found a solution by adding a data member to the property classes which
// counts the number of outstanding updtae events that it has received.  When multiple
// updates are outstanding, only propagate the last.
//
//--------------------------------
//
//When running the tests in spectest.mac, there is an error message
//QMetaMethod::invoke: Unable to handle unregistered datatype 'QVector<QString>'
//Got acquisition lock
//
//--------------------------------
//
//Save the path to the spec data file (set by newFile scripting function) and the number of columns
//to allow scans to resume if qavrg crashes;
//
//--------------------------------
//
//The 'fit' curve displayed varies wildly when the reference curve is very similar to the dark and data
//(in other words when there are no X-rays).  This may mean that the fitting algorithm is not working properly
//and should be investigated.  It might possibly account for the 'offset' problems we were ssing when making
//actual measurements.  The effect is also visible in simulation mode.
//
//Had forgotten to subtract dark data when calculating fit data
//
//--------------------------------
//
//There is clearly still a crashing problem when taking data.  I am accumulating crash dumps in the hope
//of finding out what is going on.  It would be valuable to see if the problem is also present in simulation
//mode.
//At least at first glance, it would seem that the problem is in the real acquisition mode, and doesn't affect
//simulation mode.
//Not the case - seems to be caused by a general lack of locking in the acquire classes when they are accessing their data.
//I've added many more locked sections, hopefully this will fix the crashing.  The extra locking makes it harder for
//the fitters to run concurrently - it would definitely better for each fitter channel to have it's own data.
//
//--------------------------------
//
//Should check the 'parameters adapted' error returns from the acqiris library and see if any of these might
//account for the crashing problem above.  In any case, it would be desirable to handle them gracefully.
//Probably not the problem - see above.
//
//--------------------------------
//
//Check handling of spec scan numbers - are they set anywhere
//The saved file name, scan number and file path should be made properties and replace the fileIndex,
//filePattern etc. properties which are currently unused.
//
//--------------------------------
//
//Deadlock when:
//QavrgResultsModel::data:54 reads from QavrgAcquisition::readResult:638 reads from QavrgFitter::readResult:71
//while
//QavrgAcquisitionSimulated::calculateSimulatedAcquisition:217 waiting for fitters to complete
//and
//QavrgFitter::performCalculation:302 calls QavrgAcquisitionData:set_FitData:211
//
//Need to go through the acquisition classes more carefully and eliminate unneccesary locks.
//
//--------------------------------
//
//Use orbit/bunch notation for column names in saved files
//
--------------------------------

Optimize graph update by suppressing updates until all curves are updated.
Try to get time for display of all data below 0.5 secs

--------------------------------

Bunch number choice control in 'view' panel doesn't work properly.

--------------------------------

Provide a way for the newFile command to strip off any directory components of
the path that it is passed (if it is an absolute path)

--------------------------------

Provide a 'Save Default Settings' command

--------------------------------

During scanning, it is confusing that the fitted data curves are drawn after the fitting is complete, whereas
the acquired data curves are drawn sooner.  This gives a misleading impression that the fit doens't match
up to the data.

--------------------------------

Fitting is still rather slow

--------------------------------

Scope mode can too easily overwhelm the GUI with update events.

--------------------------------

Make acquireDark & acquireRef harder to do accidentally.

--------------------------------

Crashes if acquisition.channels = 2 ?

//--------------------------------
//
//Changing number of bunches in N-bunch mode requires restart to take effect
//
--------------------------------

#endif // TODO_H