Skip to contents

Changelog

dv.loader 2.0.0

  • GitHub release with QC report
  • Update package documentation

dv.loader 1.1.1

  • General package maintenance.

dv.loader 1.1.0

  • Bugfix release that disables problematic dv.loader::load_data file name partial matching.

dv.loader 1.0.0

First release

dv.loader 0.5.1 (with breaking changes)

  • Major changes before first release
  • Changed name of package from “dataloader” to “dv.loader” to avoid CRAN collision
  • Changed the domains paramater name throughout to file_names
  • Changed the study_dir paramater name throughout to sub_dir, because it is used to reference a sub directory of get_cre_path.
  • Improved tests and test coverage
    • code coverage > 93% and refactored tests to use BDD

dv.loader 0.5.0

  • Refactored to take out the OOP design. Now you just use the functions directly without having to create a “dataloader” object.
  • Took out “meta” object from the return of load_data(). Now, metadata is appended directly to each dataframe as an attribute.
    • E.g., attr(df, "meta") to view.

dv.loader 0.4.1

  • package style changed from camelCase to snake_case
  • better documentation added with more examples

dv.loader 0.4.0

  • The API for dataloader has changed considerably. Now there are three public functions:

    • loadData() for bringing data into memory
    • createDB() for creating an SQLite database
      • now creates a new “attr” attributes table containing the column-level attributes of the files read keyed on table (domain) name.
    • getTableRefs() for retrieving a list of table references from a DB connection

  • Support for indexing the SQLite DB added.

    • Pass a list to the “file_names” arg in createDB() such that:
      • the elements of the list are domain names
      • the values of the elements are character vectors representing columns that you want indexed
      • check the result in the “sqlite_stat1” metatable retrieved from getTableRefs()

dv.loader 0.3.1

  • If you specify useDB = T and there’s an existing database file, then load_data() will connect to it, and return its tables.

  • load_data() will check for an existing DB based on the users input for dbFileName. If that is left NULL, then it will use the default name based on studyDir, and if that is NULL, then it will use db. Otherwise, it will just create a new database (if useDB = T).

  • The API to read_file() has changed. No longer need to specify isRDS. The function will figure that out based on the file_name passed in.

dv.loader 0.3.0

  • data.loader is now called dataloader

  • dataloader’s local DB functionality is now passing unit tests, meaning that the integrity of data flow from the producing system (CARE) is ensured.

  • The output of load_data() for databases includes attributes from the original file. So, even if those attributes are lost when the data is loaded into the database, you can recover them. Here is how to recover for a given column and dataframe:

attr <- dataList[["attr"]]
attr_adsl <- attr["adlb"]
attr_adsl_studyid <- attr[["adlb"]][["STUDYID"]]
attr_adsl_studyid
# $label
# [1] "Study Identifier"
#
# $format.sas
# [1] "$"

  • load_data() by default returns a list named by the file_names passed, and containing a dataframes, along with metadata for that table.

  • Create a DB table connection (created with dplyr::tbl(dl$dbConn, “myDomain”)) by using useDB = T in load_data().

  • set_base_path() has been removed, and the base_path attribute is now private so as to restrict where on CARE this module can access.

  • the only arg that is mandatory is file_names. If studyDir is left as null, then it just uses working directory as default.

dv.loader 0.2.1

  • data.loader now supports the creation of local SQL databases for managing very large files. Users should interact with the db_conn connection using the DBI package, found on CRAN. Usage is detailed below in the benchmarking section. Also, internally, the load_data() function has been refactored to separate out the importing of data from the creation of the local database.

  • This dv.loader also fixes a bug where if a user did not provide a “/” in front of the studyDir arg, then it wouldn’t be able to find the right path.

  • isLocal in the load_data() API has been replaced with useWD (for “use working directory”) to make more sense.

  • A flag for prefering SAS files over RDS files has been added to load_data()

dv.loader 0.2.0

  • data.loader is now an R6 class for internal scoping of base_path. This way, users can change the “working directory” of the data loader module without affecting the working directory of their global environment. A new function is available called set_base_path() for this purpose. See usage below for how to create a “dataloader” object. Otherwise, usage is the same as in V0.1.0.

dv.loader 0.1.0

  • Initial commit. data.loader has the functions load_data(studyDir, file_names) and set_base_path().