General ------- I feel I have to apologize for the incompleteness of this test. My goal was to concentrate on the commissioning needs, which I had thought were the focus of this test. I actually did not have much time to reproduce the data reduction steps in the cookbook, which in fact were focussed on calibration and imaging. Reproducing some of the steps in the cookbook on the data sets provided did in fact worked with a good stability, and accuracy. The comments I have are anyway more on the user command line, logging and plotting interfaces. Commissioning needs: -------------------- As mentioned above I thought this was the main focus of this test. The zeroth order need for commissioning is to be able to plot any parameter with respect to any other in a data set, and to be able to perform calculations using the actual values in the data set in a practical way. Basically what we have here is msplot, with about 10 plotting methods (uvcoverage, uvdist, ...). These are mainly aimed at displaying useful high-level statistics on an astronomical data set, but not quite at debugging and commissioning ALMA single-baseline data. What I think is more useful for commissioning/debugging is simple things like e.g.: - plotting total powers vs time, elevation, hour angle; - plotting visibility versus time or any time related parameter (elevation air mass), while averaging on any part of a spectral window or averaging on any number of spectral windows; - same thing, but time averaging on given time bins; - plotting visibility vs frequency while averaging on a time range; while binning the data on given frequency intervals... We also need the table tool in a working version, to be able to edit the data sets and do simple calculations using the contents of tables. This requires access to the table columns through python variables. This was not operational in the test, though I understand this is on track (on the version I tested there were some indexing problems, which looked to be understood). Installation: ------------- Not tested, as I worked at NRAO/AOC with the imager-c workstation, thus with a fully pre-installed system; I have no system running RedHatEnterprise available at IRAM so I could not try the installation on my laptop or at home. User Interface -------------- In general I find the command line interface has several problems: 1) There is a main window where you type in commands. - Some of the results appear there (eg.g cb.modelfit), others go to the logging window ('casalog'). It is not clear which will go where. - Each command is echoed in a translated form (with parentheses). Having the translated form is not necessary and takes unnecessary space in the window. The user should be familiar with either ipython of python; getting to know both forms is asking too much. Let the user choose one of them and stick with it. - The output of some commands (like e.g. the python dir() command ) gets formatted in a lengthy way (e.g. one line of output per item for dir()). 2) The logging window 'casalog' itself is not very practical. For some reason the output of a single command is split into several lines, interspersed with time messages. Looking into this is really annoying. See e.g. the output of im.summary(), or of cb.state(), given iun the cookbook. One should have a single header part with the date and time, the name of the function, and a part with the real contents. Optionnally one should have all output on the command-line window. Plotting system --------------- The plotting software, matplotlib, has some new interesting features. I find a basic speed limitation, and I'm not sure it is used in the best way. - When using a remote window (actually a X window on my laptop opened through ssh imager_c), I find the feature of clicking and dragging the plot area totally inoperative. This really only works on a display which is fast enough. - The labels size and/or fonts are not properly set. It is very difficult to have the full header on the plotting window, as it regularly overflows. In particular when attempting to have multiple subplots on the same window. I tried to play with mp.plotoptions(fontzize=0.5) but it did not help. I think this is probably a problem with properly defining what should be in the titles in the standard output of e.g. msplot; these should be defined to make a better use of screen space. - Some plots do not give units like uvdist (no amplitude or phase units). Documentation ------------- - The documentation on the tools is very scarce, often only giving the list of methods or arguments, while a single line of input for each would be already valuable. - The help command is also not fully operative e.g. help(sx) or help(run) give a name error; while the ? form works for sx. - Documentation of savefig advertises jpeg files but it does not work. (only png files seem supported). Stability: --------- The few crashes I had while testing were I believe corrected while I was present and were in fact all related to using features (like the table tool) that were not fully supported in the test.