I completed work on the test harness encrypted (HTTPS) PII test cases. The cases were easy to write, however the IDE caused drama's and stopped building the APK. After searching stackoverflow it appears the problem was caused by a corrupted project index file. I fixed the problem using the recommended method - rebuilding the project from scratch. In my case I imported the source Eclipse project into Android Studio, and applied recent changes (determined using a git-diff).
Finishing this change means I've addressed all of the main objectives in my GSoC project submission (I think).
Later in the afternoon, I created a new branch to work on user documentation improvements and start working though issues in the project issues list: https://github.com/mkenne11/nogotofail-pii/issues
I am done with the feature of creating diagram to diagram connections. I am working now on diagram to content connection creation. This should be straight forward; I will be using an auto-complete module I found earlier. The module suggests nodes (node titles) to the user as the user types characters. On the other side the user selects diagram elements by clicking on them and finally connects both sides (diagram elements and nodes)
I continued working on and looking into the Kalman filters tonight. Also of interest, apparently the CMSIS provides some basic matrix math functions, which was a pleasant surprise, and I plan to use them, if I can get them working.
I started to implement synchronization algorithm. I also implemented several features for MIDI Action dialog window (e.g. filtering actions) and fixed some bugs.
For now I add new MIDI Actions by editing midi-actions.xml or score file. I will have to create another dialog window for user-friendly adding MIDI event.
I figured how to turn the google drive link into a direct download link, so it downloads the file instead of pulling up a the html page to view the document. I can parse the shared link to pull the file ID and apply the id to the url pattern that will directly download the file. My sample file will download successfully using this new link. However, when i parse the url after I apply urlretrieve(), I see the instance of the file so I know that it's been created, it doesn't show the name or file type. I just get numbers where the file name & type would be (as in the case of the dropbox download). So i can't validate if a text file has been downloaded or prevent a jpg file from being downloaded into the application. I'm going to try to get this working, and I'm still working on a pop-up dialogue box that takes in an direct download link URL.
I had more luck today, and quickly found the issue causing the Android test harness to crash. The application was trying to fetch and use test PII values before they were available. Improved error messaging assisted in finding this problem.
I didn't have a chance to do much coding today, the business of the last few days caught up with me. Tomorrow I will cleanup the test harness code, I hope to do some refactoring to make the application more manageable.
I am about to finish the feature of creating diagram to diagram connections. I should finish this in an hour or two. Next is finalizing the feature of creating diagram to content (drupal nodes) connections. I am also still have minor things to do this week before the 17th due date. I will use next week for more testing, cleaning, and reviewing code.
Since the edits I'm doing are sweeping enough to make the code non-functional while the Kalman filter is being developed, I have created a separate branch in git to hold the kalman-enabled code. I have reasonable-seeming values populating most of the initial Kalman state, except for Q which is still to be determined. The next major step will be to implement the Kalman equations in the predict and update functions. Since I believe the covariance between the three axes would be zero, I plan to run three separate multivariate Kalman filters. Each filter has two hidden variables exposed in it, namely position and velocity. I also plan to implement Kalman filtering on the angular data as well, but I've started with the acceleration data.
TextMessageFilter. This allows for gRPC clients to filter and optionally modify user text messages before they are broadcast to the server.
textmessage filtercommand to murmur-cli. It will execute an external program and, based off of the external program's return value, determines whether or not the message should be rejected.
Tomorrow: wish list feature.
As I wrote to Bart Massey before, I was in a study trip, so I was busy for about two weeks.
Now I am ready to finish "MIDI Actions" feature. I came up with an algorithm for synchronizing the internal MIDI Actions in MuseScore and MIDI Actions in the opened score. We need this synchronization to have a bunch of MIDI Actions available to use in all scores. If you open a new score, new MIDI Actions will be added to MuseScore and you can use them later.