Consider promotion of integer -> double tokenizers

[kevinushey]

For certain numeric columns, it's possible that the first entries are representable as integer, with only the later entries being representable as double. In such a case, calls to read_csv() can fail as the inferred integer tokenizer will fail to tokenize those later entries.

Would it be possible for readr to instead just promote the tokenizer from integer to double on the fly in such a case?

[jimhester]

We no longer guess integer columns by default, so I am closing this.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/