R notebook push fails - script is treated as python.

I'm using the cli tool to push an r notebook to my kaggle account. The initial push fails with this error:

> 'File "", line 17
> , 'bib.refs_id' = "BIBREF0"
> ^
> SyntaxError: keyword can't be an expression'
>

The relevant source block is:
`
##A small but growing library of utilities for tidying the metadata, parallel processing
##to grow a citation table, summary of STM model build times & sizes, etc
source(
'/kaggle/usr/lib/cord_19_challenge_cord_19_utils/cord_19_challenge_cord_19_utils.R')

tibble(
cord_uid = "000ajevz"
,source = "Late‐onset Pneumocystis jirovecii pneumonia post–fludarabine, cyclophosphamide and rituximab: implications for prophylaxis"
, ref_id = "b0"
, title = "Toxicity of fludarabine and cyclophosphamide with or without rituximab as initial therapy for patients with previously untreated mantle cell lymphoma: results of a randomised phase II study"
, year = "2009"
, venue = "Leuk Lymphoma"
, volume = "50"
, issn = NA_character_
, pages = "211--216"
, other_ids = NA_character_
, 'bib.refs_id' = "BIBREF0"
, authors = "H E Eve;D Linch;W Qian;M Ross;J F Seymour;P Smith;L Stevens;Saj Rule"
)
`

The Language setting shows as R, if I manually set it to R the script then runs fine. It looks as though the notebook is assumed to be python on the initial push.

The process I've followed is:

1. Convert the .Rmd file to .ipynb using:
`jupytext cord-19-tidy-citations-1-20k-uids.Rmd --to notebook`
2. Convert the .ipynb file to .irnb format by concatenating the individual source blocks for each cell. The file is then identical to the .irnb obtained by 'pull'-ing the kernel.
3. Write the kernel-metadata.json as follows:
`{
"id": "thatsplumb/cord-19-tidy-citations-1-20k-uids",
"title": "CORD-19 Tidy Citations 1-20k-uids",
"code_file": "cord-19-tidy-citations-1-20k-uids.irnb",
"language": "r",
"kernel_type": "notebook",
"is_private": true,
"enable_gpu": false,
"enable_internet": true,
"keywords": [],
"dataset_sources": [],
"kernel_sources": [
"thatsplumb/cord-19-build-citations-table-1-10000",
"thatsplumb/cord-19-build-citations-table-10001-20000",
"thatsplumb/cord-19-challenge-cord-19-utils"
],
"competition_sources": []
}
`
4. Push the kernel.

I like the cli tool a lot, it would be great if I didn't have to reset the language to get the script to run after the push. I'm using multiple kernels to generate an aggregated citations table from the 70k + docs in the cord-19 dataset and ideally I'd like to set things up so I can re-build the table simply via a push over the cli tool.

Any idea what I'm doing wrong?

Thanks!