googleAuthR | Google API Client Library for R. Easy authentication | REST library

You can upload data frame with a list-type column on BigQuery by using bq_table_upload() syntax. Try this on your script instead of bq_table_create(),

I tried reproducing your code and I noticed that R code is not properly translated to BQ query.

1. Use sd(.x) since by default BQ (STTD_DEV) ignores the null values.
2. The function quantis is not created in BQ, thus it does not do its job and errors out. I'm not sure if R supports use of functions to BQ.

What I could suggest is instead of using native R operations use SQL statements to prevent incorrect translation of R to BigQuery operations. You can try creating a user defined function (your quantis function) in BQ. In your select statement perform mean, max and sd on your fields. Filter using your UDF(quantis), group by age and espac. You can also try creating a VIEW that achieves most of your goal including WHERE clause on quantile. You can refer to this document for reference on how to use BigQuery in R.

Take a look at this:

Use googleAuthR::gar_gce_auth() to get a token from a machine running on the Google Cloud

The R session should be waiting for you to paste it in when you switch back from your browser - see image below

But also see the googleAnalyticsR website’s setup page as this auth flow can happen automatically if you have library(httpv) installed.

Does your Dockerfile name is Dockerfile (first letter uppercase)?

Windows sometimes saves the file internally as .txt.

A Dockerfile must be created with no extension. To do this in Windows, create the file with your editor of choice, then save it with the notation "Dockerfile" (including the quotes).

Is your Dockefile in the right path at the same level as the app.yaml?

A few suggestions that may help.

• Would start with a simplified shiny app before adding all of the calculations, may be easier to work with for now
• Would avoid putting output statements inside of eventReactive. See below for example.
• Consider having only one observeEvent or eventReactive for the button press instead of multiple, especially since some function results depend on others.
• Right now there are no inputs, so don't need additional reactive expressions. When you add inputs, though, you probably will.

If you haven't already, review the R Studio Shiny tutorial on Action Buttons and Reactivity.

Hope this is helpful in moving forward.

As @G5W posted in comments, gc() after rm(df) will free up memory.

Error is due to deprecation of: auth <- callModule(gar_auth_js, "auth")

Instead the following must be used. auth <- callModule(googleAuth_js, "auth")