A selectizeInput() below the current textInput could have its choices consist of the titles of already cached shows, as to give suggestions on which show data too look at (and also giving an indicator of which shows have been searched for in a deployed instance).

However, since currently only the episode datasets are cached and their filenames consist of tvdb_ids, it would require to greatly expand or rewrite the current caching process, which, to be fair, is pretty basic as of now.

Sometimes the show title / summary doesn't update,
sometimes the plot doesn't render until updated.

I don't know what's going on, but it's bad.

Fix it

Should probably use checkboxInput, with some kind of

…ggvis stuff…
if (checkbox.value == TRUE){
scale_numeric… %>%
}
…

However, the %>%-piping doesn't work through if statements, so there will probably be the need for ggplot2-like statements in the form of

plot <- ggvis…
plot <- plot %>% morestuff
plot <- plot %>% evenmorestuff
return(plot)

¯_(ツ)_/¯

EDIT1: Come to think of it, the latter form of combining ggvis elements is probably a good idea anyway, since we'll probably have a few more plot options as user inputs. The more modular approach seems more appropriate in this case than the serial, straightforward piping mechanism.

First up: The way deployment to shinyapps.io works is that everything the application writes to disk does not survive consecutive deployments, i.e. the vm the app runs in gets destroyed and a new one is created. Therefore caching is even more temporary that one would assume.

However, I think we should create a cache dir, and every series dataset that gets pulled should be saved in the most efficient way (either via save or saveRDS, which would probably be more flexible) to said directory, and on startup the app should be able to compare the query (or download the search result and then compare id_tvdb or something) and if the dataset is present in cache, read the data from there.

We could also download some popular series' data and permanently commit it to the repository, as example data. This would probably be helpful for debugging purposes as well.

Happens whenever I try to Ploterizzle any show.

When are the cached shows refreshed?
The walking dead seems to be two weeks behind in the plotted graphic for example.

Related issue: trakt/api-help#33

This is the reason “Presidio Med” shows up a bunch in the cache, presumably because people were looking for the show "24".

Well ain't I glad I finally found out some search result weirdness.

Create function that takes the episode dataset as an argument, and returns the dataset with a new column including the html-tooltip for each episode.

It should contain the following columns' values at least:

• epid
• title
• firstaired.string
• rating

And probably also:

• votes
• id.tvdb

Which would need to be formatted appropriately and combined with <br />-tags because html 'n stuff.

Should be possible via layer_smooth() or something.
Probably also easiest to use checkboxInput again on this one.

something like

ggvis() %>%
  ...
  handle_click(utils::browseURL("http://www.r-project.org")) %>%
  ...

should/could somehow provide clickable data-points.

I.e. for example /?show=True+Detective should insert True Detective in the textInput (as it does now) but also trigger a data update, which currently solely relies on the value of the actionButton being >= 0.
The goal would be to make it easier to share certain results.

Since the query would be evaluated when the actionButton is still at zero, it could be used as an additional argument to conditions that check if actionButton == 0.

Or something like that.