8 Settings and conditions

8.1 Options and environment variables

Setting	Kind	Effect
`croissant.cache_dir`	R option	Default `cache_dir` of `read_record_set()` and `hf_download()`; otherwise `tools::R_user_dir("croissant", "cache")`
`HF_TOKEN`	Environment variable	Hugging Face token when the `token` argument is `NULL`
`HF_ENDPOINT`	Environment variable	Hugging Face Hub endpoint; otherwise `https://huggingface.co`

8.2 Cache layout

Directory	Content
`<cache_dir>/downloads/<hash>/`	Files downloaded from `http(s)` URLs, keyed by their `sha256`, their `md5`, or a hash of the URL
`<cache_dir>/extracted/<hash>/`	Extracted zip and tar archives, keyed by a hash of the archive’s path, size, and modification time
`<cache_dir>/hf/datasets/<owner>/<name>/<commit>/`	Files downloaded from Hugging Face datasets

Deleting a cache directory removes the copies; the next read downloads or extracts the files again.

8.3 Conditions

Each error class below inherits from rtemis_input_error and rtemis_error, so a handler for either class catches all of them.

Class	Raised when
`croissant_read_error`	`read_croissant()` finds problems that prevent reading; the `issues` element holds a `CroissantIssues` object
`croissant_version_error`	A document is Croissant 0.8 or conforms to a version other than 1.0 or 1.1
`croissant_json_error`	The input is not valid JSON, or is JSON but not an object
`croissant_structure_error`	A document holds more than one top-level node after JSON-LD processing
`croissant_file_error`	A file cannot be found, located, or extracted, or has an unsupported `contentUrl` scheme
`croissant_download_error`	A download fails
`croissant_checksum_error`	A downloaded file does not match its checksum
`croissant_format_error`	A field extracts a column from a file whose encoding format is not a supported table format
`croissant_record_error`	A record set cannot be assembled, for example because of a missing column or an undeclared join
`croissant_jsonpath_error`	A JSONPath uses syntax that `croissant` does not evaluate
`rtemis_dependency_error`	A suggested package required for the operation is not installed
`rtemis_file_error`	The input file does not exist
`rtemis_type_error`	An argument has the wrong type, such as a column class that `croissant_from_data()` does not map
`rtemis_value_error`	An argument has an invalid value
`rtemis_unsupported_error`	A gated Hugging Face dataset is requested without a token

tryCatch(
  read_record_set(path, "records"),
  croissant_checksum_error = function(e) {
    message("A file does not match its checksum: ", conditionMessage(e))
  }
)

8.4 API reference

The API reference documents every exported function and class.