fhir_search() (e.g. with rm_tag = "div") now get caught and converted to a warning.fhir_collapse() collapses multiple entries that belong to the same higher level FHIR element and should be kept together, e.g. name.given or address.line
fhir_melt_all() divides all multiple entries into separate rows by repeatedly calling fhir_melt().
Thanks to @atruebi and @Flow191 who contributed the code for these functions!
fhir_crack() can now deal with XPath expressions using predicates, e.g. code/coding[system[@value='http://loinc.org']]/codefhir_remove_empty_cols fully removedas_fhir() coerces a character vector with xml strings representing FHIR bundles to a fhir_bundles_list object.fhir_url() now properly formats requests containing =, e.g. when searching for snomed expressions in coding elementstry() to stop them from crashing during CRAN checksfhir_search() now prints the correct base URL in downloading messages even when http:// type CodeSystems occur in search params.fhir_search() has a new argument add_headers that allows adding custom headers to the HTTP request.fhir_load() can now load FHIR bundles with arbitrary file names from a given directory.fhir_search() will not url encode next links for paging anymore, as servers are expected to deliver a URL that can be sent back to them without any preprocessing steps. On all tested server types (HAPI, Vonk, Blaze) this change of code doesn't affect the paging behavior in any way.fhir_search() with token authentication have been fixed.fhir_authenticate() doesn't cache a .httr-oauth file anymore.fhir_url() now url encodes string type parameter values too.fhir_search() can now log errors from FHIR servers regardless of whether they are valid xml.fhir_search() can now handle relative next links when paging through bundlesfhir_build_bundle(): Build a xml FHIR Bundle from a wide table.fhir_build_resource(): Build a xml FHIR resource from a wide table.fhir_bundle_list(): Constructor function to create fhir_bundle_list objects is now exported.fhir_cast(): Similar to fhir_melt() but spreads multiple entries across columns instead of rows.fhir_count_resource(): Count resources matching a specific search request on a server.fhir_get_resource_ids(): Extract the logical IDs of all FHIR resources matching a specific search request from a server.fhir_get_resources_by_ids(): Extract a set of resources based on their logical IDs.fhir_post(): POST an object to a FHIR server.fhir_put(): PUT an object to a FHIR server.fhir_recent_http_error(): Returns the most recent http error to the console.fhir_request(): Wrapper for fhir_url()fhir_resource_xml(): Constructor function to create fhirresource_xml() objects is now exported.fhir_rm_div(): Remove html elements from FHIR bundles.fhir_rm_tag(): Remove elements enclosed by a specified xml-tag from FHIR bundles.fhir_sample_resources_by_ids(): Randomly sample resources from a given logical ID vector and download them from a FHIR server.fhir_sample_resources(): Randomly sample from a set of resources matching a specific search request in a FHIR server.fhir_tree(): Returns a string representing the tree structure implicit in the column names of a wide table.pastep(): Concatenates two or more strings to a path string correctly.fhir_melt() has been rewritten to be faster and more memory efficient.
fhir_style: This class is now deprecated, all the information contained in the fhir_style object has been moved to the fhir_table_description.
fhir_table_description:
fhir_table_description has changed insofar that the elements that used to be part of fhir_style are now direct elements of fhir_table description.format and keep_attr. For more information see ?fhir_table_descriptionfhir_crack():
ncores > 1.wide format where multiple entries are distributed over several columns instead of being pasted together in one column (argument format).@value, @id, @url), attributes can be attached to the column names when argument keep_attr = TRUE.remove_empty_columns renamed to rm_empty_cols.fhir_search():
New argument rm_tag="div" per default removes all html bits in the bundle immediately after download. Can be disabled by setting rm_tag=NULL.
delay_between_attempts can now be a vector with different waiting times between retries of reaching the server. The length of delay_between_attempts determines the number of retries.
max_attempts is deprecated in favor of delay_between_attempts.
fhir_load() and fhir_save() have a new argument max_bundles that allows to restrict the number of bundles that is loaded/saved.
The package is now rewritten using S4 classes. Almost all of the code written with fhircrackr < 1.0.0 will still work, though in a few cases the user will be prompted to change their code to the new syntax.
The most prominent change is how to create a design that tells fhir_crack() how to flatten HL7 FHIR® Resources. It should now be created using the functions fhir_table_description() and fhir_design(). fhir_crack() is now also able to create a single data.frame/data.table instead of a list of tables when just one resource type is extracted. Please see the documentation under ?fhir_table_description and ?fhir_design!
To get an overview about how the general workflow has changed, please have a look at the intro vignette (vignette("fhircrackr_intro", package="fhircrackr")).
There are a couple of new constructor functions for the newly defined classes which will not be listed here. Please have look at the package vignettes which go through them in detail.
Other new functions or behavior are listed in the following:
fhir_authenticate(): Set up authentication using OAuth2/OpenID Connectfhir_current_request(): Get search request used in most recent call to fhir_search()or fhir_url()fhir_search() now allows for a search via POST via the argument body.
fhir_search() can now handle bearer token authentication via the argument token.
Argument save_to_file in fhir_search() now takes NULL or a string with a directory name and saves the bundles only if there is a specified directory. For backwards compatibility TRUE/FALSE in combination with directory are still allowed but discouraged with a warning.
Argument log_errors in fhir_search() now takes a string with a filename and writes an xml (no tables anymore) to the specified file. For backwards compatibility numbers are still allowed but discouraged with a warning.
New argument delay_between_bundles for fhir_search() allows to put a delay between the download of bundles (i.e. pages) in a bigger search request to prevent weak servers from choking.
The output of fhir_capability_statement() is slightly restructured: The names of the tables and their structure has changed.
The deprecated argument add_indices in fhir_crack() is now fully removed.
The most recently used FHIR search request is now implicitly saved whenever fhir_search() or fhir_url() is called. It can be accessed with the new function fhir_current_request().
The default value of rm_empty_cols is now FALSE
fhir_melt now gives a warning when the brackets provided in the function call don't appear in the data frame
A number of bugs have been fixed
fhir_melt() now takes the name provided in the argument id_name (which because of a bug it didn't before)fhir_crack() have been clearedcols element of design was of length onedesign for fhir_crack() has new form now:
has now named elements resource, cols, style (with style elements sep, brackets, rm_empty_cols)
old versions of design still work
new function fhir_canonical_design() returns the full (potentially automatically completed) design of the most recent call to fhir_crack()
argument add_indices of fhir_crack() is now deprecated, indices will be added when brackets is not NULL
new argument columns of fhir_rm_indices() gives control over the columns in which indices should be removed
new functions fhir_save_design() and fhir_load_design() for saving/loading design as xml-document
new function fhir_next_bundle_url() returns next-link of the last bundle processed by the most recent call to fhir_search()
new arguments save_to_disc and directory of fhir_search() allow for saving bundles consecutively as xml files instead of loading them into the R session all at once
Faster results of fhir_crack() because it now uses data.table internally
new argument data.table of fhir_crack to choose between data.frame vs. data.table as output format
fixed errors in fhir_crack() when resource type doesn't appear in bundle
handle errors caused by the accidental use of serialized objects more gracefully
@value at the end of an XPath expression pointing to an attribute for design used in fhir_crack() is now optional and will be added automatically by fhir_crack if omitted
column names automatically generated by fhir_crack() are now shorter.
First Release of R-Package fhircrackr v0.1.0