Resilience to unexpected interactions

Various interactions I can have with the interface (to be clear, deliberately silly interactions, like just clicking some text elements) cause a hard crash, or in the case of the hosted version a "disconnected from server" error. An example:

From a clean R installation:

install.packages("devtools")
library(devtools)
devtools::install(".", dependencies = TRUE)
library(viewpoly)
viewpoly::run_app()
## click on the "Input data" tab
## click on the text "Upload linkage map files"

I've found a half dozen other interactions that hard crash like this. I understand that the interactions are not intended ones, but it seems like some of those might better be caught and handled in a more graceful way. This is really a matter of user convenience, I guess: having to relaunch everything because I happened to click on the text instead of the "+" is a bit of a bummer.

I'm not really sure from a dev perspective how much of a nightmare it would be to prevent this kind of behavior in general. But for some of the text elements, for example, it seems like clicking on them when no other interactions have occurred might reasonably result in nothing in particular happening, as opposed to a hard crash and a minimal stack trace.

[Cristianetaniguti]

Hello @cpalmer718 ! Thanks for your feedback. I made a silly mistake in this particular error that you mentioned (forgot to import a package). Now, clicking in "Upload linkage map files" or any box title should do the same thing that the + sign does.
I will keep fixing this type of error once I become aware of them.

Hi @Cristianetaniguti yes, I see the text click works now, that's great. Here's a few other interactions with strange results:

(1) steps to reproduce:

install.packages("devtools")
library(devtools)
devtools::install(".", dependencies = TRUE)
library(viewpoly)
viewpoly::run_all()
## click on tab "Input data"
## click on each of the "reset" buttons in the subsections
## click the "Download" button under "Download VIEWpoly dataset"
## the download prompt suggests a file named "upload_ui_1-export_viewpoly" which lacks the .RData extension and seems somewhat malformed

(2) steps to reproduce:

• same installation/launch steps as above
• in the "Upload VIEWpoly dataset" section, select the option "There is no viewpoly object in your R environment. Load view viewpoly object or convert formats below
• click "submit VIEWpoly file"
• hard crashes, emits the following:

Warning: Error in get: object 'There is no viewpoly object in your R environment. Load view viewpoly object or convert formats below' not found
  [No stack trace available]
Warning: Error in get: object 'There is no viewpoly object in your R environment. Load view viewpoly object or convert formats below' not found
  50: <Anonymous>
Warning: Error in get: object 'There is no viewpoly object in your R environment. Load view viewpoly object or convert formats below' not found
  50: <Anonymous>

(3) steps to reproduce:

• go to QTL view
• select a view that leads to "Error: no individual in the progeny was found containing the combination of all the selected homolog/s. Please, select another combination of homolog/s."
• click on "Download"
• suggested filename is "qtl_view_ui_1-bn_download_haplo", same kind of strange extension/malformation issue (requested file format in this example was ".png")

(4) same issue as above, except:

• select a view that leads to valid combination

• set width/height/DPI to 0

• suggested filename is again "qtl_view_ui_1-bn_download_haplo"

• seems like some version of this strange filename thing happens whenever the output is invalid; maybe in such situations the download could instead be disabled? or whatever other handler seems appropriate.

[Cristianetaniguti]

Hi @cpalmer718 ,

(1) This was the only that I couldn't reproduce here. Does it display any other error message in the console? Anyway, the download will be disabled if the object is empty (see below).

(2) I changed the error with a warning. Now, if the user performs these steps the app doesn't crash, but a message is displayed in the console and in the graphics box in the next tabs.

(3) and (4) I thought that disabling the button in these situations was a really good idea, thanks! I figure out that it is possible to do that with shinyjs package functions.

@Cristianetaniguti well, deeply mysteriously, I'm getting variable behavior. The first time I tried to reproduce (1) again with upstream patches installed, I got a well-behaved .RData file. So that was encouraging. I then reloaded the app and tried again. This time, the download prompt took a moment to process, and in my general scattered state I alt-tabbed to a different window and switched back a moment later. The download prompt appeared, but for some reason asked if I wanted to download a file called qwe_download.html. I downloaded it; it's attached, if you're curious; it also came along with some sort of bizarrely hashed file that I've included in the attachment for posterity. Again, this is just from the same download dataset button. So I shut down the app and tried the alt-tabbing mid-download thing again. This time, I managed to completely freeze both the app and R, and had to kill the process.
example_strange_files.zip

I don't know, this sounds like weird GUI nonsense here. I won't block your review based on something like this, it's clearly a bigger issue with shiny or something, and these were admittedly strange behaviors; and it worked the first time. You can check it out if you feel like it, but if you can't reproduce, there's really nothing to be done.

Thanks for checking out the things I mentioned, I appreciate it. Hopefully this will prevent some user difficulties.

[Cristianetaniguti]

@cpalmer718 🤔
It is really mysteriously... I still can't reproduce it here.
Does the update that uses the disable download button avoid this behavior?

@Cristianetaniguti as far as I can tell, I have all of upstream patched and installed, but it still happens intermittently. I do think the patch helped. It seems like it starts out fine at first, but then something gets messed up in the background and it becomes a problem. I can't reliably trigger it, but it seems like after I close and reopen the app a few times, it gets stuck with the qwe_download.html option and stays there. Now even if I reboot R it consistently triggers that download. I'm not sure if I have to reboot to get the correct behavior back or what. It doesn't seem like there's any weird background zombie process or anything. All very mysterious.