koderkow / kaggler Goto Github PK

When running kgl_flow("titanic") I get the following error:

• These files will be downloaded:
  - 'gender_submission'
  - 'test'
  - 'train'.
Error in `as_tibble()`:
! All columns in a tibble must be vectors.
✖ Column `gender_submission` is a `httr2_response` object.
✖ Column `test` is a `httr2_response` object.
✖ Column `train` is a `httr2_response` object.
Run `rlang::last_trace()` to see where the error occurred.

This is the backtrace:

rlang::last_trace(drop = FALSE)
<error/tibble_error_column_scalar_type>
Error in `as_tibble()`:
! All columns in a tibble must be vectors.
✖ Column `gender_submission` is a `httr2_response` object.
✖ Column `test` is a `httr2_response` object.
✖ Column `train` is a `httr2_response` object.
---
Backtrace:
     ▆
  1. ├─kaggler::kgl_flow("titanic")
  2. │ └─v_to_download %>% ...
  3. └─purrr::map_dfr(...)
  4.   └─dplyr::bind_rows(res, .id = .id)
  5.     ├─tibble::as_tibble(dots)
  6.     └─tibble:::as_tibble.list(dots)
  7.       └─tibble:::lst_to_tibble(x, .rows, .name_repair, col_lengths(x))
  8.         └─tibble:::check_valid_cols(x, call = call)
  9.           └─tibble:::abort_column_scalar_type(...)
 10.             └─tibble:::tibble_abort(...)
 11.               └─rlang::abort(x, class, ..., call = call, parent = parent, use_cli_format = TRUE)

The flow should be able to submit data to a competition.

• I believe the API requires a flat-file to be submitted
  • Previous developer does have initial competition submit code, but I have not tested it
• If we can have the user run something like kgl_flow_submit(predictions) and have it saved to a folder, _kaggle_data/submissions that would be gnarly
  • Maybe we can store the user submissions in a programmatic way? Perhaps the name of the dataset in R, or, have an optional field "submission name" that defaults to the name of the r dataset. Perhaps always add a _01 suffix
  • If a user submits the same name dataset we would have to save it with an incrementing number
    • example: my_submission_02
• Before submitting, check user-generated predictions against the given sample submission to make sure the column structure is correct to avoid errors on Kaggle
• With the new directory, _kaggle_data/submissions/ we need to make sure the previous Kaggle Flow functions do not error when scanning for all files

Having some functions to show the leaderboard would be nice!

• I should only have to run kgl_flow_leaderboard(), this will use the stored competition_id in meta/
• Perhaps another function, kgl_flow_my_results (name up for change!), that will show the rank of the current users submissions