Is a dhdl.xvg file needed with .collect() and .analyze()?

[a-ws-m]

The docs suggest using .collect() and .analyze(), but there's a couple of issues as far as I can tell:

1. The default Gsolv.setup(), using edr=True, sets the .mdp option separate-dhdl-file = no. There are therefore no .xvg files produced.
2. The dgdl_xvg() function looks for a file whose name is <deffnm>.xvg, swapping out the bit in tags, but the default name for the file produced when using -dhdl is dhdl.xvg.

I've switched to using the alchemlyb implementation of these functions, but it would be good to know if I was doing something wrong and to update the above example if not.

PS I can see that analyze() handles the IOError thrown by dgdl_xvg() and looks for the .edr file instead. Maybe I'm not supposed to call .collect() before .analyze()?

[orbeckst]

The fact that we default to EDR and then collect() only looks for XVG seems to indicate that the docs have not been kept up-to date or that the code has simply become inconsistent. It would make more sense if collect() looked for EDR files first. The whole logic in analysis() in the block https://github.com/Becksteinlab/MDPOW/blob/120c5dddd67d14c493971125e6355414b09e73c1/mdpow/fep.py#L1089C1-L1097C26

            try:
                self.collect(stride=stride, autosave=False)
            except IOError as err:
                if err.errno == errno.ENOENT:
                    self.convert_edr()
                    self.collect(stride=stride, autosave=False)
                else:
                    logger.exception(err)
                    raise

would be better in collect(). Basically,

1. check which format files are present, XVG or EDR?
2. if EDR, convert them to xvg.bz2
3. then parse everything

In short, the easiest thing would be to remove the collect() call from the docs because analyze() does it better at the moment.

Ultimately, exactly the same thing happens with collect_alchemlyb() (can only use XVG and fails if there's none) and analyze_alchemlyb().

(Sidenote: A while back we switched to using EDR files by default because they are more compact. Unfortunately, we still do what seems a waste, namely, converting EDR free energy data to XVG because we haven't implemented an actually EDR reader (with pyedr).

It would also be really good if someone found the time to add a EDR reader to alchemlyb alchemistry/alchemlyb#67 because then we wouldn't need to do the conversion.)