Create functions to mirror dataframes into REDCap projects

[ljwoodley]

We need to mirror dataframes into REDCap.

Spec for mirror_data_to_redcap_project

• [ ] The function mirror_data_to_redcap_project take a single parameter, table_name
• [ ] The function will read the environment to get the value of the environment variable named <table_name>_pid.
• [ ] The function will also read the value of the env variables named CREDENTIALS_DB.
• [ ] The function will search CREDENTIALS_DB for the REDCap API credentials for the current user in the PID pointed to by the value of <table_name>_pid.
• [ ] The function will use REDCapR::redcap_write() to write the data frame to the REDCap project described by those credentials.
• [ ] The function will return the complete results of REDCapR::redcap_write()

Spec for dataframe_to_redcap_dictionary

If we can't find one in another R package, we should write a helper function named something like dataframe_to_redcap_dictionary. The function should output a CSV we can import into an empty REDCap project so that we could write the data of that dataframe into that project. The function should do these things:

• [ ] Transform the column names into a column of field_name
• [ ] Add a column, form_name with a value of table_name
• [ ] Add a column, section_header with a value of as.character(NA)
• [ ] Add a column field_type with a value of "text"
• [ ] Transform the column names into a column of field_label
• [ ] Add a column, select_choices_or_calculations with a value of as.character(NA)
• [ ] Add a column, field_note with a value of as.character(NA)
• [ ] Transform the column classes valid text validation types and write that to a column named text_validation_type_or_show_slider_number
• [ ] Return the constructed dataframe

If we do this, make this function part of REDCap Custodian.

For reference, see https://github.com/kamclean/collaborator/blob/master/R/data_dict.R which generates a different data dictionary despite being in a redcap-centric package. :-(

This issue was moved from https://github.com/ctsit/rcc.billing/issues/174

[ljwoodley]

1. Would mirror_data_to_redcap_project be an upsert, overwrite or append? Trying to determine how best to create the record_id.

2. Should we account for all these validation types?

date_dmy, 
date_mdy,
date_ymd, 
datetime_dmy,
datetime_mdy,
datetime_ymd,
datetime_seconds_dmy,
datetime_seconds_mdy,
datetime_seconds_ymd,
email, 
integer,
alpha_only,
mrn_generic,
number, 
number_1dp,
number_2dp,
number_3dp, 
number_4dp,
phone, 
ssn, 
time_hh_mm_ss, 
time, time_mm_ss,
zipcode 

@pbchase

[pbchase]

1. Would mirror_data_to_redcap_project be an upsert, overwrite or append? Trying to determine how best to create the record_id.

You may assume that every mirrored table has a PK (aka record_id) and that every write event to that table you are mirroring into REDCap has that PK field in it.

[pbchase]

2. Should we account for all these validation types?

date_dmy, 
date_mdy,
date_ymd, 
datetime_dmy,
datetime_mdy,
datetime_ymd,
datetime_seconds_dmy,
datetime_seconds_mdy,
datetime_seconds_ymd,
email, 
integer,
alpha_only,
mrn_generic,
number, 
number_1dp,
number_2dp,
number_3dp, 
number_4dp,
phone, 
ssn, 
time_hh_mm_ss, 
time, time_mm_ss,
zipcode 

Please support these validations:

• If you have a datetime column in the dataframe, set the validation type to datetime_seconds_ymd.
• If you have a date column in the dataframe, set the validation type to datetime_ymd.
• If the character data matches an email address pattern, set the validation type to email.
• If the class is numeric, set validation to number.

Everything else should probably be invalidated text. I say this, because REDCap is willing to do special things with columns if they are these types. The other column types are more to control data entry. I don't plan on any data entry for these forms, so there is no point in the additional complexity.