AI SC Report - AI funded software position

[sarahcleary]

Request for report on AI funded software position at each SC meeting

ACTION @Keith-Pembleton and @sarahcleary to work with @hol430 on paper (possible template to be used in future reports)

[sarahcleary]

@hol430 has forwarded a script that downloads the total number of bugs fixed.

[hol430]

The script is a work in progress - I'm working on generating more stats/charts. Might put a live version on a webpage at some point.

[hol353]

How about lines, not symbols? Increase font size.

[hol430]

How about these? Any suggestions or other stats you would like to see?

[sarahcleary]

Circulated to RP for review and comment.

[sarahcleary]

Reopen Issue - standing item for AI SC meetings

[sarahcleary]

@hol430 - next AI SC meeting is 22 November. Paper will need to be prepared by @Keith-Pembleton and @sarahcleary by 8/11.

[hol430]

What action is required from me - do you just want updated graphs?

[sarahcleary]

@hol430 - at this stage, that would be great.

[hol430]

[sarahcleary]

@hol430 - could you please redo the Total issues open and closed graph - to show the issues labelled as ‘interface/infrastructure’ - as I understand these are the ones you work on.

[hol430]

Have regenerated all graphs such that they ignore all issues which do not have the 'interface/infrastructure' label. According to my tool I've fixed 272 issues with this label and 322 issues total.

[hol353]

Bottom graph is interesting in that @hol430 has taken over from me, around middle of 2018, as chief bug fixer. Perhaps this goes into report @sarahcleary along with a sentence to that effect. Also interesting that Drew going onto 50% UQ time earlier this year has probably had an impact on the number of bugs fixed - see graph below. Dash red line is the trend of Drew's bug fixing up until he went 50:50 AI:UQ. After that the rate of bug fixing has declined. This is expected of course.

[sarahcleary]

@hol353 - When did the ‘interface/infrastructure’ tag start being utilised?

[hol430]

We applied the labels to all open issues around the end of 2017 I think, which means that this graph is a bit misleading. A more accurate comparison of our relative bug fix rates would be in the same graph in my above comment which includes total number issues closed. Also how did you generate that trend line? It doesn't look much like the long term average

[hol353]

Not sure. Probably at least a couple of years ago.

[hol353]

I just added the trend line by hand in paint. Nothing scientific. There appears to be a lag until about July 2018 when you suddenly started fixing a lot more bugs (the start of my hand drawn line). Ideally, I'd like to see a proper trend-line starting from the lag point (July 2018) up until you started on the 50:50 joint appointment and another trend line after this point.

[sarahcleary]

@hol430 - just checking - is this statement still valid? 322 bugs fixed since commencing

[hol430]

322 issues resolved is probably a more accurate statement but I guess it amounts to more or less the same thing. An issue can be a bug or a question or a feature request or some refactoring or whatever.

[sarahcleary]

@hol353 and @sno036 - will you be in a position to share your update with the RP on Tuesday?

[sno036]

There is a great visualisation of this that I'll try and show at the meeting - something I suggested but Dean solved and is based on interrogating GitHub issues. I might work for the SC.

[sarahcleary]

As per #50
Steering Committee would like guidance from RP with regard to the Software Engineering role. What resourcing is required long term. What additional tasks could/should be done. Looking at not just fixing bugs, but e.g. new features. Linked with conversation around RP role - #51

[sno036]

In reality Drew has not just been fixing bugs/defects. While he has been doing some of that, in reality he has been fixing issues which often as not means new features which make things easier/better/faster/robuster. Because he is a software person rather than a modeller (a lamentable lack of science training - what was his paternal parental unit thinking!) that is probably both the best major role for him and the thing that we most need. If, as well as improving the documentation generation / the website / the build and test procedures / running on clouds / xxx / xxx / xxx, he had some time left over to do a combination of going back into existing code and giving it a good refactor / tidy up and working alongside modellers doing some of the more tricky bits I think we would get a huge amount out of the role and end up with a much better and less frustrating product.

[sarahcleary]

@hol430 - are you able to update these graphs? and list the main issues/tasks you've been working on since Nov 2019?

[hol430]

[sarahcleary]

@hol430 - have you a list like this one for the period Nov 19-March 2020??

• Syntax highlighting for report (#4251)
• Parameter optimization from R via simplex method. This is not built into APSIM yet but is a standalone tool
• Implemented framework for automated UI tests (#4199)
• Moved APSIM.Shared into ApsimX to make compilation easier (#4121)
• Significant speed improvement for multi-paddock simulations (#4082)
• Port of old apsim’s SoilTemp2 model (#4080)

[hol430]

I haven't really fixed any major issues in the last few months - the issues I've been working on have been smaller in scope. The only big change that comes to mind is:

• Report now throws a fatal error if variables cannot be found (#4838)

[sarahcleary]

Visualisation now on APSIM youtube channel: https://www.youtube.com/watch?v=waU_ALZ589o

[sarahcleary]

Closing, will open if necessary

[sarahcleary]

@hol430 - it's that time of year, and the next AI SC meeting is on 25/3. Are you able to provide me with updated graphs and list of any major issues you've been working on.
@jbrider / @hol353 / @hut104 - anything you would like to highlight to the AI SC with regard to @hol430's work over the past 4 months?
I will stick with the standard report unless you have any other suggestions.
thanks!

[hol430]

Interesting things I've been working on:

• Reworking the CI system to be entirely cloud-based so that we can decommission our physical build servers
• Get next gen running on .net core / gtk3
• Changed the default garbage collector to server mode (this will significantly speed up cluster/HPC runs - although on linux/mac, you'll need to be running the .net core build to see any difference)
• Have contributed to some 3rd-party projects which are dependencies of apsim (mix of bug fixes and new features which we needed in apsim)
• Have added some official docker images for apsim (nextgen only so far, but we will be providing images for classic as well in the fullness of time). This will be helpful for people in hpc scenarios

[hol430]

@sarahcleary - as requested:

Number of open issues:                      312
Number of closed issues:                    2435
Number of open pull requests:               0
Number of closed pull requests:             3859
Number of bugs closed since 1/1/1970:       867
Number of issues closed since 1/1/1970:     2435
Number of issues fixed by hol430:           722

"Major" issues:

• Faster opening/closing of files in the GUI (Speed up the opening of files in the GUI ApsimX#6563, Refactor of command system in ui ApsimX#6239)
• APSIM Server (Added APSIM.Server project ApsimX#6530). This server will open a file and hold it in memory, running the file on-demand with minimal latency. The idea is it will be used by optimisers/sensitivity analyses, which need to rerun the model many times. Previously the runtime has been swamped by the slow opening of files and other overheads associated with repeatedly invoking Models.exe.
• Allow post-simulation tools to be run in parallel (Add the ability to run post-simulation tools in parallel ApsimX#6578)

[sarahcleary]

@hol430 - can you please provide updated graphs and any new major issues? thanks

[hol430]

Number of open issues:                              343
Number of closed issues:                            2569
Number of open pull requests:                       0
Number of closed pull requests:                     3994
Number of bugs closed since 1/1/1970:               929
Number of issues closed since 1/1/1970:             2569

• Refactor autodocs project so that it doesn't depend on gtk (this has allowed us to retire our last physical server)
• Remove last traces of .net framework (reduces development/maintenance overheads)