wiki:AppsWGMeeting20131009
Last modified 5 years ago Last modified on 10/09/2013 11:53:59 AM

2013-10-09

Present: JFB, REO, PAP, RAA, KSK, ACB, Perry, Yusra, AJC, KTL, RHL

RHL: The results of the survey are in!  Sorted by mean rank (the individual responses are at the bottom).
IDDescription Mean Rank
10. Work on documentation 3.9
4. Perform need major code audit (decide what was good, what needs a rewrite) including ticket scrub 5.4
12. Upgrade package and distribution tools 5.8
6. Rewrite/upgrade Butler 6.4
13. Upgrade/replace buildbot and continuous integration and testing 7.0
3. Rewrite/design CameraGeom 7.8
2. Rethink overall code design and Python/C++ (including Tasks) 7.9
11. Tools to look at data (pipeQA update, debug plotting, get rid of DS9) 8.5
16. Generate unified python/C++ docs (using sphinx+doxygen bridge) 9.3
1. Redo source measurement framework 9.9
9. Change image coordinate conventions to (almost?) always use PARENT 10.2
8. Refactor Source association/matching into a callable and scalable module 10.5
14. Move development to stash, mirror on github; move dev.lsstcorp.org to Confluence/Jira 10.8
5. Rethink to see if we can integrate with AstroPy world 11.5
15. Switch to C++11 12.0
7a. Rewrite logging 12.4
7b. Replace policy 13.0
RHL: In the absence of the D.M., let's start at the bottom.

RHL: Log/Policy is last
KTL: Jacek wants to do this [as part of qserv?] in early construction.

RHL: Policy will wither on the vine (camGeom/Mapper).
KTL: also pex_harness

RHL: C++11
REO: I'd like to use auto.
KTL: The main issue is whether we convert the entire codebase.
JFB: The real issue is std::unique_ptr
REO: Do we need a new sconsUtils?
JFB: Done, but clang on os/x had problems.
KTL: As soon as everything compiles, we'll permit at least some C++11;  we'll need a style guide and suggestions.
RHL: Who should we task with this guide?  JFB: Let's start with more-or-less, "Only use auto"

RHL: AstroPy is becoming the de facto standard
JFB: We should start by including it in our stack.  There's some overlap (wcs objects; table objects) and we could use their versions where appropriate
KTL: Do they have interfaces that we could code to?
JFB: Not really.  They're pretty laisez-faire.
KTL: What's the goal?
RHL: Make LSST accessible to astroPy world.
JFB: Maybe just provide mappings between the worlds (e.g. method on our Wcs to make it look like an astroPy one)
KTL: Not a top-down rethink

RHL: The advantage of stash is the code review support
JFB: Not so important for new users.
PAP: It'll take us some time to learn the new systems.  We should become comfortable ourselves.
AJC: The project hasn't signed off yet.
KTL: It's official, but not at Victor's level.  Where's the money?  Who'll set it up?
RHL: So Apps should tell DM to push for it; we're ready

RHL: What depends on source association/matching?
JFB: Looking at data does
AJC: The problem is that it isn't taskified.
RHL: Do we care about AP, isn't it just database ingest?
KTL: No!  It's apps code
AJC: Let's ask the people who use it, Simon and Yusra
Yusra: It took an awful long time..
RHL: It's way downstream of most of the apps code
KSK: We just did forced photometry ... Yusra: but now we need to associate in SDSS
KTL: The uses are matching difference sources with objects, and SFM with objects.
RHL: This is all downstream of where we're working now; 

RHL: XY0 (PARENT/CHILD).
JFB: Not very big deal, but infinitely confusing to newcomers
RHL: Is this really that small?
JFB: I think the scary thing is the Image accessors in C++
RHL: This wasn't ranked highly, does this mean that we don't care?
Perry: It's very confusing.
KTL: Should be fixed, but we're scared.

RHL: Source measurement framework
JFB: This mostly affects me, but not many others.
KSK: I thought it would be disruptive, which was why I up-voted it

RHL: Do people care about python/doxygen docs?
REO: It's crazy not to do it
JFB: It's more important to get the high-level stuff right
RHL: Don't the python docs get lost?
REO: I thought that this just worked for the camera team...
KSK: I don't use the web-based docs (I look at the code or pydoc), so this is more important for newbies?

RHL: Tools to look at data.  Can we afford a complete FTE?
JFB: We need part of an FTE
RHL: It's ranked 8
Perry: Isn't this just
ACB:  I'm not sure this is in R4, but it's really important -- but not needed pre-new-people
RHL: Could it be a starter project?
PAP: I disagree.  We need this to let new people make progress on the code
ACB: There are tools already.
AJC: How bad are the tools?  How much time will this waste?  I agree with PAP
JFB: It slows us down less than it will new people.
PAP: It'll be a particular problem for new science-type hires.
JFB: Yes, but the best way to write this sort of thing is to work on particular problems, while thinking about the framework.
     I'd like the luxury to work on the framework and tools simultaneously.
RHL: But wouldn't a study on what we have done already, followed by a design

RHL: We're out of time.  Fortunately the rest of the topics seem better defined.
KTL: We need a 2-4 month breakdown of people's work packages by this afternoon for FDR...
RHL KTL SRP KSK RAA PAP REO ACB Perry Yusra Mean
10. Work on documentation 3 7 3 3 4 3 9 2 4 1 3.9
4. Perform need major code audit (decide what was good, what needs a rewrite) including ticket scrub 10 3 1 5 5 10 1 1 13 5 5.4
12. Upgrade package and distribution tools 8 10 2 8 1 5 2 6 13 3 5.8
6. Rewrite/upgrade Butler 1 5 10 10 12 1 5 13 1 6 6.4
2. Rethink overall code design and Python/C++ (including Tasks) 4 1 10 12 6 9 16 4 13 4 7.9
13. Upgrade/replace buildbot and continuous integration and testing 5 4 10 9 2 6 12 7 13 2 7.0
3. Rewrite/design CameraGeom 2 12 10 4 12 7 4 13 2 12 7.8
11. Tools to look at data (pipeQA update, debug plotting, get rid of DS9) 11 13 10 7 3 4 8 13 3 13 8.5
16. Generate unified python/C++ docs (using sphinx+doxygen bridge) 7 8 10 14 12 8 11 3 5 15 9.3
1. Redo source measurement framework 17 11 10 1 12 13 6 13 6 10 9.9
9. Change image coordinate conventions to (almost?) always use PARENT 13 14 10 2 12 11 3 13 13 11 10.2
8. Refactor Source association/matching into a callable and scalable module 12 15 10 6 12 14 7 13 7 9 10.5
14. Move development to stash, mirror on github; move dev.lsstcorp.org to Confluence/Jira 14 9 10 13 12 2 13 8 13 14 10.8
5. Rethink to see if we can integrate with AstroPy world 6 2 10 17 12 15 15 13 8 17 11.5
15. Switch to C++11 9 17 10 16 12 12 10 5 13 16 12.0
7a. Rewrite logging 15 6 10 15 12 16 17 13 13 7 12.4
7b. Replace policy 16 16 10 11 12 17 14 13 13 8 13.0