UTF-8

[hadley]

Should check that on windows (which doesn't use UTF-8) standardise_csv() converts to UTF-8 on write, and maintains UTF-8 encoding status on read

[krlmlr]

I guess a more robust approach would be to use JSON for uploading instead of CSV. jsonlite seems to work fine with UTF-8 on Windows (jeroen/jsonlite#42). Another advantage: I've run into problems with corner cases such as a one-row one-column data frame with an empty value, this probably won't be an issue when JSON is used.

[hadley]

And now worth considering if we should use DML: https://cloud.google.com/bigquery/docs/reference/standard-sql/data-manipulation-language (but probably slower than json)