Create a dm object from data frames

2021-06-13

Building data models from data frames with dm

dm allows you to create your own relational data models from local data frames. Once your data model is complete, you can deploy it to a range of DBMSs using dm.

Creating a dm object from data frames

The example data set that we will be using is available through the nycflights13 package. The five tables that we are working with contain information about all flights that departed from the airports of New York to other destinations in the United States in 2013:

Once we’ve loaded {nycflights13}, the aforementioned tables are all in our work environment, ready to be accessed.

library(nycflights13)
#> Error in library(nycflights13): there is no package called 'nycflights13'

airports
#> Error in eval(expr, envir, enclos): object 'airports' not found

Your own data will probably not be available as an R package. Whatever format it is in, you will need to be able to load it as data frames into your R session. If the data is too large, consider using dm to connect to the database instead. See vignette("howto-dm-db") for details on using dm with databases.

Adding Tables

Our first step will be to tell dm which tables we want to work with and how they are connected. For that we can use dm(), passing in the table names as arguments.

library(dm)

flights_dm_no_keys <- dm(airlines, airports, flights, planes, weather)
#> Error in .f(.x[[i]], ...): object 'airlines' not found
flights_dm_no_keys
#> Error in eval(expr, envir, enclos): object 'flights_dm_no_keys' not found

The as_dm() function is an alternative that works if you already have a list of tables.

A dm is a list

dm objects behave like lists with a user- and console-friendly print format. In fact, using a dm as a nicer list for organizing your data frames in your environment is an easy first step towards using dm and its data modeling functionality.

Member referencing, by subscript and by name, and list and slice manipulation functions work just as you would expect on a dm object.

names(flights_dm_no_keys)
#> Error in eval(expr, envir, enclos): object 'flights_dm_no_keys' not found
flights_dm_no_keys$airports
#> Error in eval(expr, envir, enclos): object 'flights_dm_no_keys' not found
flights_dm_no_keys[c("airports", "flights")]
#> Error in eval(expr, envir, enclos): object 'flights_dm_no_keys' not found

Defining Keys

Even though we now have a dm object that contains all our data, we have not specified how our five tables are connected. To do this we need to define primary keys and foreign keys on the tables.

Primary keys and foreign keys are how relational database tables are linked with each other. A primary key is a column or column tuple that has a unique value for each row within a table. A foreign key is a column or column tuple containing the primary key for a row in another table. Foreign keys act as cross references between tables. They specify the relationships that gives us the relational database. For more information on keys and a crash course on databases, see vignette("howto-dm-theory").

Primary Keys

dm offers dm_enum_pk_candidates() to identify viable primary keys, and dm_add_pk() to add them.

dm_enum_pk_candidates(
  dm = flights_dm_no_keys,
  table = planes
)
#> Error in is_dm(dm): object 'flights_dm_no_keys' not found

Now, we add the primary keys that we have identified:

flights_dm_only_pks <-
  flights_dm_no_keys %>%
  dm_add_pk(table = airlines, columns = carrier) %>%
  dm_add_pk(airports, faa) %>%
  dm_add_pk(planes, tailnum)
#> Error in is_dm(dm): object 'flights_dm_no_keys' not found
flights_dm_only_pks
#> Error in eval(expr, envir, enclos): object 'flights_dm_only_pks' not found

Note that we demonstrate both named and positional arguments above.

Foreign Keys

To define how our tables are related, we use dm_add_fk() to add foreign keys. In calling dm_add_fk() the table argument is the table that needs a foreign key to link it to a second table. ref_table is the table to be linked to and it needs a primary key already defined for it.

dm_enum_fk_candidates(
  dm = flights_dm_only_pks,
  table = flights,
  ref_table = airlines
)
#> Error in is_dm(dm): object 'flights_dm_only_pks' not found

Having chosen a column from the successful candidates provided by dm_enum_fk_candidates(), we use the dm_add_fk() function to establish the foreign key linking the tables. In the second call to dm_add_fk() we complete the process for the flights and airlines tables that we started above. The carrier column in the airlines table will be added as the foreign key in flights.

flights_dm_all_keys <-
  flights_dm_only_pks %>%
  dm_add_fk(table = flights, columns = tailnum, ref_table = planes) %>%
  dm_add_fk(flights, carrier, airlines) %>%
  dm_add_fk(flights, origin, airports)
#> Error in is_dm(dm): object 'flights_dm_only_pks' not found
flights_dm_all_keys
#> Error in eval(expr, envir, enclos): object 'flights_dm_all_keys' not found

Having created the required primary and foreign keys to link all the tables together, we now have a relational data model we can work with.

Visualization

Visualizing a data model is a quick and easy way to verify that we have successfully created the model we were aiming for. We can use dm_draw() at any stage of the process to generate a visual representation of the tables and the links between them:

flights_dm_no_keys %>%
  dm_draw(rankdir = "TB", view_type = "all")
#> Error in is_dm(dm): object 'flights_dm_no_keys' not found

flights_dm_no_keys %>%
  dm_add_pk(airlines, carrier) %>%
  dm_draw()
#> Error in is_dm(dm): object 'flights_dm_no_keys' not found

flights_dm_only_pks %>%
  dm_add_fk(flights, tailnum, planes) %>%
  dm_draw()
#> Error in is_dm(dm): object 'flights_dm_only_pks' not found

flights_dm_all_keys %>%
  dm_draw()
#> Error in is_dm(dm): object 'flights_dm_all_keys' not found

Integrity Checks

As well as checking our data model visually, dm can examine the constraints we have created by the addition of keys and verify that they are sensible.

flights_dm_no_keys %>%
  dm_examine_constraints()
#> Error in is_dm(dm): object 'flights_dm_no_keys' not found

flights_dm_only_pks %>%
  dm_examine_constraints()
#> Error in is_dm(dm): object 'flights_dm_only_pks' not found

flights_dm_all_keys %>%
  dm_examine_constraints()
#> Error in is_dm(dm): object 'flights_dm_all_keys' not found

The results are presented in a human-readable form, and available as a tibble for programmatic inspection.

Programming

Helper functions are available to access details on keys and check results.

dm_get_all_pks() returns a data frame with our primary keys:

flights_dm_only_pks %>%
  dm_get_all_pks()
#> Error in is_dm(dm): object 'flights_dm_only_pks' not found

A data frame of foreign keys is retrieved with dm_get_all_fks():

flights_dm_all_keys %>%
  dm_get_all_pks()
#> Error in is_dm(dm): object 'flights_dm_all_keys' not found

We can use tibble::as_tibble() on the result of dm_examine_constraints() to programmatically inspect which constraints are not satisfied:

flights_dm_all_keys %>%
  dm_examine_constraints() %>%
  tibble::as_tibble()
#> Error in is_dm(dm): object 'flights_dm_all_keys' not found