Open judfs opened 1 month ago
Once again I dared to use unicode. -encoding utf8
made the tests pass at least.
I like intent of this PR. It seems a lot more sane to skip writing out a .jlp
if the user does nothing but click the GUI window to close it.
One of the things I liked about the old behavior was this:
But in this PR, a .jlp
isn't saved unless I disable a channel. What do you think about preserving the above behavior?
My motivation for this change is that it is inconvenient for additional files to be created when just playing back a log and doing nothing fancy. My team shares datasets as folders of LCM logs and random jlp files are undesired clutter. To this end, I would like not creating a jlp file when simply playing a log to be the default. "But in this PR, a .jlp isn't saved unless I disable a channel. What do you think about preserving the above behavior?": That was my entire goal.
I'd be interested in a opt in "Always make jlp" user preference.
I'd also be interested in a button to create a more proper index file for a log. I can imagine recording each of the following per channel:
logplayer-gui could then populate the filterlist from an index.
@judfs what are your thoughts about keeping the original behavior as the default, but adding a flag that enables the behavior you suggested above?
My preference is to keep the original behavior as the default to provide UX stability for users who have come to expect that behavior, but I definitely agree that there are use cases where the default is inconvenient. We could also consider environment variables instead of command line flags.
How about a dialogue on close?
Save log-player-gui session (.jlp)? [Yes] [No] Remember this for future logs [ ]
In a side-channel, we discussed this a bit more. In all, the solutions we've considered are:
.nojlp
sentinel/marker file. If such a file exists alongside the log then the logplayer-gui will not write a .jlp.
If you just open and close a log, without clicking on any knobs, don't create a
.jlp
file.google-java-format
was suggested at https://github.com/lcm-proj/lcm/issues/420#issuecomment-1472374172 . I made a commit using it on the file I was working with first. Looking at the diff of 36a670eee9217152c2a65301d3dd69ba05d2e9f8 should be sane.