These changes could possibly break existing code --but it's very unlikely. We feel they will (directly and indirectly) improve the package considerably.
redcap_read(), redcap_read_oneshot(), redcap_dag_read(), redcap_log_read(), and redcap_report() return a tibble instead of a data.frame. (#415)
This should affect client code only if you expect a call like ds[, 3] to return a vector instead of a single-column data.frame/tibble. One solution is to upcast the tibble to a data.frame (with something like as.data.frame()). We recommend using an approach that works for both data.frames and tibbles, such as ds[[3]] or dplyr::pull(ds, "gender").
For more information, read the short chapter in R for Data Science.
The *_collapsed parameters are deprecated. When your want to limit on records/fields/forms/events, pass the vector of characters, not the scalar character separated by commas (which I think everyone does already). In other words use c("demographics", "blood_pressure") instead of "demographics,blood_pressure".
Here are the relationships between the four pairs of variables:
records_collapsed <- collapse_vector(records , records_collapsed)
fields_collapsed <- collapse_vector(fields , fields_collapsed)
forms_collapsed <- collapse_vector(forms , forms_collapsed)
events_collapsed <- collapse_vector(events , events_collapsed)
If someone is using the *_collapsed parameter, they can programmatically convert it to a vector like:
field_names <- trimws(unlist(strsplit(field_names_collapsed, ",")))
redcap_read() will automatically include the "plumbing" variables, even if they're not included the list of requested fields & forms. (#442). Specifically:
record_id (or it's customized name) will always be returnedredcap_event_name will be returned for longitudinal projectsredcap_repeat_instrument and redcap_repeat_instance will be returned for projects with repeating instrumentsThis will help extract forms from longitudinal & repeating projects.
redcap_read() and redcap_read_oneshot() now return an empty dataset if no records are retrieved (such as no records meet the filter criteria). Currently a 0x0 tibble is returned, but that may change in the future. Until now an error was deliberately thrown. (#452)
redcap_event_instruments() now by default returns mappings for all arms. The previous default was to return the mappings for only the first arm. To recreate the previous behavior use a call like REDCapR::redcap_event_instruments(uri, token_2, arms = "1"). (Suggested by @januz, #482)
redcap_users_export() used to return a boolean for user_rights, but now it can be 0, 1, or 2. (#523)
redcap_metadata_coltypes() function. Inspects the fields types and validation text of each field to generate a suggested readr::col_types object that reflects the project's current data dictionary. The object then can be passed to the col_types parameter of redcap_read() or redcap_read_oneshot(). (Suggested and discussed with @pbchase, @nutterb, @skadauke, & others, #405 & #294)redcap_log_read() function. Exports a project's log. (Thanks @joundso, #383, #320)redcap_project_info_read() function. Exports a project's information, such as its language and production status. (Suggested by @skadauke, @timothytsai, @pbchase, #236, #410)blank_for_gray_form_status in the functions redcap_read(), redcap_read_oneshot(), and redcap_read_oneshot_eav(). (@greg-botwin, #386, #389)httr::handle value is accepted by functions that contact the server. This will accommodate some institutions with unconventional environments. (Suggested by @brandonpotvin, #429)sanitized_token() now accepts an alternative regex pattern. (Suggested by @maeon & @michalkouril, #370)redcap_read_eav_oneshot() is an UNexported function that returns data in an EAV format (#437)redcap_metadata_read() now correctly subsets the forms (identified & corrected by @ezraporter, #431 & #445)redcap_event_read() function. Exports metadata associated with a project's longitudinal events (@ezraporter, #457 & #460)Better documentation for the server url (suggested by @sutzig #395)
read_read_oneshot()'s parameter guess_max now allows floating point values to support readr::read_csv() ability to accept a Inf value. (Suggested by @eveyp, #392)
pkgdown pages run & display the examples, but CRAN still doesn't run them. It's illegal to call external resources/APIs from CRAN computers --mostly because they are occasionally unavailable, so the code breaks. (#419)
Renamed some functions to follow a consistent pattern. Old names will be soft-deprecated for a while before being removed. (#416)
redcap_download_file_oneshot() to redcap_file_download_oneshot()redcap_file_upload_oneshot() to redcap_file_upload_oneshot()redcap_download_instrument() to redcap_instrument_download()redcap_dag_read() has new data_access_group_id field (introduced maybe in 13.1.0) (#459)
redcap_users_export() has new mycap_participants field (introduced maybe in 13.0.0) (#459)
Accommodate older versions of REDCap that don't return project-level variable, like has_repeating_instruments_or_events, missing_data_codes, external_modules, bypass_branching_erase_field_prompt (@the-mad-statter, #465, #466)
redcap_meta_coltypes() correctly determines data type for autonumber record_id fields. It suggests a character if the project has DAGs, and an integer if not. (@pwildenhain, #472)
redcap_log_read() now returns a new column reflecting the affected record id value (ref #478)
redcap_read() and redcap_read_oneshot() now remove "pseudofields" (e.g., redcap_event_name, redcap_repeat_instrument, & redcap_repeat_instance) from the fields parameter. Starting with REDCap v13.4.10, an error is thrown by the server. REDCap will return a message if a common pseudofield is requested explicitly by the user. (#477)
redcap_event_instruments() now can return mappings for all arms, instead of one arm per call.(Suggested by @januz, #482)
validate_for_write() contains a few more checks. (#485) The complete list is now:
validate_data_frame_inherits()validate_field_names()validate_record_id_name()validate_uniqueness()validate_repeat_instance()validate_no_logical()redcap_read() checks the event parameter and throws an error if a value is not recognized, or the project is not longitudinal (#493)
The regex in regex_named_captures() is forgiving if there's an unnecessary leading space (@BlairCooper, #495, #501)
redcap_log_read() assumes all columns are character, except for timestamp (#525)
redcap_file_download_oneshot() no longer asks for the unnecessary parameter for repeating_instrument (that the REDCap server ignores). (@BlairCooper, #506, #530)
redcap_read() and redcap_read_oneshot() accommodate readr::read_csv()'s parameter of na. (Suggested by @rmtrane in #529)
redcap_delete() function. It deletes a vector of records. (Thanks @joundso, #372, #373)redcap_arm_export() function. It retrieves a list of REDCap project arms. (#375)redcap_read() and redcap_read_oneshot() accept a new locale parameter that specifies date, time, and number formats, like using a comma as the decimal separator. It is a readr::locale object. (#377, suggested by @joundso)redcap_instruments() function exports a list of the data collection instruments for a project. (#381, @vcastro)redcap_event_instruments() function exports the instrument-event mappings for a project (i.e., how the data collection instruments are designated for certain events in a longitudinal project). (#381, @vcastro)redcap_dag_read() function returns the Data Access Groups for a project (#382, @joundso)sanitize_token() now allows lowercase characters --in addition to uppercase characters & digits. (#347, @jmbarbone)redcap_metadata_read() now uses json (instead of csv) to transfer the dictionary between server & client. This accommodates super-wide dictionaries with 35k+ variables. The user shouldn't notice a difference, and still will receive a data.frame. (#335, @januz & @datalorax)testthat::skip_on_cran() calls to comply with CRAN's "fail gracefully" policy. Similarly, skip remaining examples that depend on external resources. (#352)retrieve_credential_local() can now user username to identify the desired credential row (#364)redcap_read() and redcap_read_oneshot() gain the http_response_encoding parameter that's passed to httr::content(). The default value remains "UTF-8". (#354, @lrasmus)redcap_users_export() (which calls REDCap's user export). The API dropped the data_export variable and added the forms_export variable. (#396)redcap_read_oneshot_eav(): if the project isn't longitudinal, a dummy value for event_id is used internally (#396)httr::user_agent, following the advice of httr's vignette (#397)The package has been stable for years and should be reflected in the major version number.
redcap_write() and redcap_write_oneshot() have a new parameter that converts R's logical/boolean columns to integers. This meshes well with T/F and Y/N items that are coded as 1/0 underneath. The default will be FALSE (ie, the integers are not converted by default), so it doesn't break existing code. (#305)redcap_write() and redcap_write_oneshot() can toggle the ability to overwrite with blank/NA cells (suggested by @auricap, #315)redcap_read_oneshot(), redcap_read(), & redcap_read_oneshot_eav() now support the parameters datetime_range_begin and datetime_range_end. The are passed to the REDCap parameters dateRangeBegin and dateRangeEnd, which restricts records returned, based on their last modified date in the server. (Thanks @pbchase, #321 & #323.)export_survey_fields parameter in the functions redcap_read() & redcap_read_oneshot(). (Thanks @isaactpetersen, #333)redcap_report() export records that populate a REDCap report. (#326.)create_credential_local() starts a well-formed csv file that can contain tokens. (#340, after conversations with @higgi13425 & @kamclean.)config_option element. (Proposed by @BastienRance, #307)success value returned by redcap_read() and redcap_write() when the parameter continue_on_error is true. (Bug found by @llrs, #317)redcap_survey_link_export_oneshot() documentation corrected & improved (@jrob95, #526)kernel_api() defaults to "text/csv" and UTF-8 encoding. Formerly, the function would decide on the content-type and encoding. More details are below in the 'Stability Features' subsection.
constant() no longer accepts simplify as an options. An integer vector is always returned. (#280)
It's now possible to specify the exact col_types (a readr::cols object) that is passed to readr::read_csv() inside redcap_read_oneshot(). (#258)
reader::type_convert() is used after all the batches are stacked on top of each other. This way, batches cannot have incompatible data types as they're combined. (#257; thanks @isaactpetersen #245) Consequently, the guess_max parameter in redcap_read() no longer serves a purpose, and has been soft-deprecated. (#267)
redcap_metadata_write() writes to the project's metadata. (#274, @felixetorres)
redcap_survey_link_export_oneshot() retrieves the URL to a specific record's survey (e.g., "https://redcap-dev-2.ouhsc.edu/redcap/surveys/?s=8KuzSLMHf6") (#293)
convert_logical_to_integer is a new parameter for redcap_write() and redcap_write_oneshot(). If TRUE, all [base::logical] columns in ds are cast to an integer before uploading to REDCap. Boolean values are typically represented as 0/1 in REDCap radio buttons. Defaults to FALSE to maintain backwards compatibility. (#305)
httr::content() (which is inside kernel_api()) now processes the returned value as "text/csv", by default. This should prevent strange characters from tricking the process as the internal variable raw_text is being formed. See the [httr::content()](https://httr.r-lib.org/reference/content.html) documentation for a list of possible values for the content_type` parameter. (Thanks to great debugging by @vortexing #269, @sybandrew #272, & @begavett, #290)
Similarly, kernel_api() now has an encoding parameter, which defaults to "UTF-8". (#270)
'checkmate' package is now imported, not suggested (Thanks @dtenenba, #255).
Allow more than one httr::config() parameter to be passed (Thanks @BastienRance, #307).
redcap_next_free_record_name(): API call for 'Generate Next Record Name', which returns the next available record ID (Issue #237)redcap_read() and redcap_read_oneshot() allow the user to specify if all variables
should be returned with the character data type. The default is to allow readr::read_csv() to guess the data type. (#194)redcap_read_oneshot() allows use to specify how many rows should be considered when readr::read_csv() guesses the data type. (#194)redcap_read(), redcap_read_oneshot(), and redcap_read_oneshot_eav() always return Linux-style line endings (ie \n) instead of Windows style line endings (ie, \r\n) on all OSes. (#198)read_metadata() always returns character vectors for all variables. With readr 1.2.0, some column were returned differently than before. (#193)redcap_users_export() now included (#163)redcap_read(), redcap_read_oneshot(), & redcap_read_oneshot_eav()(#206). It was already implemented for redcap_metadata_read().constants()) easily exposes REDCap-specific constants. (#217)id_position allows user to specify if the record_id isn't in the first position (#207). However, we recommend that all REDCap projects keep this important variable first in the data dictionary.redcap_read() and redcap_read_oneshot() are more consistent with the order in raw REDCap API. (#204)verbose parameter is NULL, then the value from getOption("verbose") is used. (#215)guess_max parameter provided in redcap_read() (no longer just redcap_read_oneshot()). Suggested by @isaactpetersen in #245.redcap_variables() now throws an error when passed a bad URI (commit e542155639bbb7).kernal_api() function, which uses the 'httr' and 'curl' packages underneath. Until now, each function called those packages directly. (#213)readr::read_csv() is used instead of utils::read.csv() (Issue #127).odbc package to retrieve credentials from the token server. Remove RODBC and RODBCext (#188). Thanks to @krlmlr for error checking advice in https://stackoverflow.com/a/50419403/1082435.data.table::rbindlist() replaced by dplyr::bind_rows()testit::assert() and base:stop() ) (#190 & #208).collapse_vector() is refactored and tested (#209)pkgload package (#218)retrieve_token_mssql(), because retrieve_credential_mssql() is more general and more useful.redcap_read() and redcap_read_oneshot() (PR #126)retrieve_credential_mssql() and retrieve_credential_local(). These transition from storing & retrieving just the token (ie, retrieve_token_mssql()) to storing & retrieving more information. retrieve_credential_local() facilitates a standard way of storing tokens locally, which should make it easier to follow practices of keeping it off the repository.data.table::rbindlist() is used. This should prevent errors with the first batch's data type (for a column) isn't compatible with a later batch. For instance, this occurs when the first batch has only integers for record_id, but a subsequent batch has values like aa-test-aa. The variable for the combined dataset should be a character. (Issue #128 & https://stackoverflow.com/questions/39377370/bind-rows-of-different-data-types; Thanks @arunsrinivasan)dplyr package instead of plyr. This shouldn't affect callers, because immediately before returning the data, REDCapR::redcap_read() coerces the tibble::tibble (which was formerly called dplyr::tbl_df) back to a vanilla data.frame with as.data.frame().redcap_variables().redcap_read_oneshot_eav(), which can be accessed with a triple colon (ie, REDCapR::redcap_read_oneshot_eav()).retrieve-token() tests now account for the (OS X) builds where the RODBC package isn't available.curl package, instead of RCurl).requireNamespace() instead of require().readcap_read() is being used without 'Full Data Set' export privileges. The problem involves the record IDs are hashed.id_position in the first stage of batching. The metadata needed to be read before that, after the updates for REDCap Version 6.0.x.retrieve_token_mssql() uses regexes to validate parametersconfig_options in the httr package are exposed to the REDCapR user. See issues #55 & #58; thanks to @rparrish and @nutterb for their contributions (https://github.com/OuhscBbmc/REDCapR/issues/55 & https://github.com/OuhscBbmc/REDCapR/issues/58).redcap_metadata_read() are tested and public.testthat::skip_on_cran() before any call involving OUHSC's REDCap server.redcap_write() and redcap_write_oneshot() are now tested and public.redcap_write() and redcap_write_oneshot() are now tested and public.redcap_download_file_oneshot() function contributed by John Aponte (@johnaponte; Pull request #35)redcap_upload_file_oneshot() function contributed by @johnaponte (Pull request #34)verbose==TRUE. Follows advice of @johnaponte, Benjamin Nutter (@nutterb), and Rollie Parrish (@rparrish). Closes #43.records_collapsed default empty value is now an empty string (i.e., "") instead of NULL. This applies when records_collapsed is either a parameter, or a returned value.redcap_download_file_oneshot() documentation, thanks to Andrew Peters (@ARPeters #45).httr package, which provides benefits like the status code message can be captured (eg, 200-OK, 403-Forbidden, 404-NotFound). See https://cran.r-project.org/package=httr.status_message to outcome_message. This is because the message associated with http code returned is conventionally called the 'status messages' (eg, OK, Forbidden, Not Found).raw_text value (which was formerly called raw_csv) is returned as an empty string to save RAM. It's not really necessary with httr's status message exposed.redcap_column_sanitize() function to address non-ASCII charactersredcap_write() (as an internal function).redcap_project() object reduces repeatedly passing parameters like the server URL, the user token, and the SSL cert location.redcap_read_batch() to redcap_read(). These changes reflect our suggestion that reads should typically be batched.redcap_read() to redcap_read_oneshot()redcap_write() to redcap_write_oneshot() (which is an internal function).redcap_read() and redcap_read_batch() with documentationredcap_read() takes parameter for raw_or_label (Thanks Rollie Parrish #3)redcap_read() takes parameter for export_data_access_groups thanks to Rollie Parrish (@rparrish #4)GitHub Commits and Releases