NEWS

---

oenb 0.1.1

Improvements

• The new help page ?oenb describes how a query is built, from finding the ID of a data set with oenb_toc to downloading the data with oenb_data, and how the results of oenb_attributes and oenb_frequency map onto the arguments of oenb_data.
• The help page of each function lists the columns of its result.
• The messages and errors of the package carry classes, so that code can tell the reasons for a NULL or an empty result apart without matching their text: oenb_unavailable, oenb_unparsable and oenb_no_results for messages, and oenb_service_error and oenb_invalid_argument for errors. They inherit from oenb_message and oenb_error, respectively. All but oenb_invalid_argument carry the URL of the request. The text of the messages is unchanged, except that the messages of oenb_attributes, oenb_frequency and oenb_metadata now point to oenb_dataset when nothing is found.

oenb 0.1.0 (2026-09-18)

Bug fixes

• The check of the lang argument never had any effect, so that an unsupported language was silently passed on to the web service. All functions now raise an error for anything other than "DE" and "EN".
• oenb_data failed with object 'temp_pos' not found whenever a query did not return any data, for example for a valid series outside its available period. It now returns NULL with an informative message.
• oenb_metadata failed with names do not match previous names for every indicator whose metadata contain a nested element that holds a single entry, such as a data_available block that describes one period. In a sample of the available series roughly four out of ten were affected. Nested elements are now skipped consistently, whether they hold one entry or several.
• oenb_data arranged the attribute columns by sorting their names, so for an indicator with ten or more attributes attr10 was placed between attr1 and attr2 and an attribute was separated from the column describing it. The value under each column name was always correct, so code that selects columns by name was not affected; only code that relies on their position was. The columns are now arranged by number.
• oenb_frequency reports which frequencies can be chosen for a series, but listed every one of them once per combination of the attributes of that series, because that is how the web service reports them. For some indicators this returned tens of thousands of entries, in one case 28512 of them for three frequencies. Every choice is now listed once, as oenb_attributes already listed the values of an attribute once. Frequencies that differ are kept, and indicators that were not affected return what they returned before.
• Arguments are percent-encoded before they are inserted into a query, so that values containing reserved characters no longer produce a malformed URL.

Improvements

• All functions now fail gracefully with an informative message and return NULL if the web service cannot be reached, as required by the CRAN policy on packages that use internet resources.
• Error messages of the web service, for example about an unknown position code, are passed on to the user instead of surfacing as internal errors such as subscript out of bounds.
• Queries that return no match now yield an empty data frame with the documented columns instead of failing.
• Added a test suite based on recorded responses of the web service, so that the package can be tested without an internet connection.

Other changes

• The minimum required R version is now 4.1.0, which is the version required by the imported package dplyr.
• Added a citation file, so that citation("oenb") returns a reference for the package instead of one generated from the description.

oenb 0.0.2 (2021-03-22)

• Added a NEWS.md file to track changes to the package.
• Adapted oenb_data to deal with a recent change of the API output