Class dm and basic operations

2021-06-20

The goal of the {dm} package and the dm class that comes with it, is to make your life easier when you are dealing with data from several different tables.

Let’s take a look at the dm class.

Class dm

The dm class consists of a collection of tables and metadata about the tables, such as

All tables in a dm must be obtained from the same data source; csv files and spreadsheets would need to be imported to data frames in R.

Examples of dm objects

There are currently three options available for creating a dm object. The relevant functions for creating dm objects are:

  1. dm()
  2. as_dm()
  3. new_dm()
  4. dm_from_src()

To illustrate these options, we will now create the same dm in several different ways. We can use the tables from the well-known {nycflights13} package.

Pass the tables directly

Create a dm object directly by providing data frames to dm():

library(nycflights13)
#> Error in library(nycflights13): there is no package called 'nycflights13'
library(dm)
dm(airlines, airports, flights, planes, weather)
#> Error in .f(.x[[i]], ...): object 'airlines' not found

Start with an empty dm

Start with an empty dm object that has been created with dm() or new_dm(), and add tables to that object:

library(nycflights13)
#> Error in library(nycflights13): there is no package called 'nycflights13'
library(dm)
empty_dm <- dm()
empty_dm
#> dm()
dm_add_tbl(empty_dm, airlines, airports, flights, planes, weather)
#> Error in list2(...): object 'airlines' not found

Coerce a list of tables

Turn a named list of tables into a dm with as_dm():

as_dm(list(airlines = airlines,
           airports = airports,
           flights = flights,
           planes = planes,
           weather = weather))
#> Error in as_dm(list(airlines = airlines, airports = airports, flights = flights, : object 'airlines' not found

Turn tables from a src into a dm

Squeeze all (or a subset of) tables belonging to a src object into a dm using dm_from_src():

sqlite_src <- dbplyr::nycflights13_sqlite()
#> Error in (function (cond) : error in evaluating the argument 'drv' in selecting a method for function 'dbConnect': there is no package called 'RSQLite'

flights_dm <- dm_from_src(sqlite_src)
#> Error in dm_from_src(sqlite_src): object 'sqlite_src' not found
flights_dm
#> ── Metadata ───────────────────────────────────────────────────────────────
#> Tables: `airlines`, `airports`, `flights`, `planes`, `weather`
#> Columns: 53
#> Primary keys: 4
#> Foreign keys: 4

The function dm_from_src(src, table_names = NULL) includes all available tables on a source in the dm object. This means that you can use this, for example, on a postgres database that you access via src_postgres() (with the appropriate arguments dbname, host, port, …), to produce a dm object with all the tables on the database.

Low-level construction

Another way of creating a dm object is calling new_dm() on a list of tbl objects:

base_dm <- new_dm(list(trees = trees, mtcars = mtcars))
base_dm
#> ── Metadata ───────────────────────────────────────────────────────────────
#> Tables: `trees`, `mtcars`
#> Columns: 14
#> Primary keys: 0
#> Foreign keys: 0

This constructor is optimized for speed and does not perform integrity checks. Use with caution, validate using validate_dm() if necessary.

validate_dm(base_dm)

Access tables

We can get the list of tables with dm_get_tables() and the src object with dm_get_src().

In order to pull a specific table from a dm, use:

tbl(flights_dm, "airports")
#> Warning: `tbl.dm()` was deprecated in dm 0.2.0.
#> Use `dm[[table_name]]` instead to access a specific table.
#> # A tibble: 86 x 8
#>    faa   name                     lat    lon   alt    tz dst   tzone       
#>    <chr> <chr>                  <dbl>  <dbl> <dbl> <dbl> <chr> <chr>       
#>  1 ALB   Albany Intl             42.7  -73.8   285    -5 A     America/New…
#>  2 ATL   Hartsfield Jackson At…  33.6  -84.4  1026    -5 A     America/New…
#>  3 AUS   Austin Bergstrom Intl   30.2  -97.7   542    -6 A     America/Chi…
#>  4 BDL   Bradley Intl            41.9  -72.7   173    -5 A     America/New…
#>  5 BHM   Birmingham Intl         33.6  -86.8   644    -6 A     America/Chi…
#>  6 BNA   Nashville Intl          36.1  -86.7   599    -6 A     America/Chi…
#>  7 BOS   General Edward Lawren…  42.4  -71.0    19    -5 A     America/New…
#>  8 BTV   Burlington Intl         44.5  -73.2   335    -5 A     America/New…
#>  9 BUF   Buffalo Niagara Intl    42.9  -78.7   724    -5 A     America/New…
#> 10 BUR   Bob Hope                34.2 -118.    778    -8 A     America/Los…
#> # … with 76 more rows

But how can we use {dm}-functions to manage the primary keys of the tables in a dm object?

Primary keys of dm objects

Some useful functions for managing primary key settings are:

  1. dm_add_pk()
  2. dm_get_all_pks()
  3. dm_rm_pk()
  4. dm_enum_pk_candidates()

If you created a dm object according to the examples in “Examples of dm objects”, your object does not yet have any primary keys set. So let’s add one.

We use the nycflights13 tables, i.e. flights_dm from above.

dm_has_pk(flights_dm, airports)
#> [1] TRUE
flights_dm_with_key <- dm_add_pk(flights_dm, airports, faa)
#> Error: Table `airports` already has a primary key. Use `force = TRUE` to change the existing primary key.
flights_dm_with_key
#> Error in eval(expr, envir, enclos): object 'flights_dm_with_key' not found

The dm now has a primary key:

dm_has_pk(flights_dm_with_key, airports)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found

To get an overview over all tables with primary keys, use dm_get_all_pks():

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

Remove a primary key:

dm_rm_pk(flights_dm_with_key, airports) %>%
  dm_has_pk(airports)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found

If you still need to get to know your data better, and it is already available in the form of a dm object, you can use the dm_enum_pk_candidates() function in order to get information about which columns of the table are unique keys:

dm_enum_pk_candidates(flights_dm_with_key, airports)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found

The flights table does not have any one-column primary key candidates:

dm_enum_pk_candidates(flights_dm_with_key, flights) %>% dplyr::count(candidate)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found

dm_add_pk() has a check argument. If set to TRUE, the function checks if the column of the table given by the user is unique. For performance reasons, the default is check = FALSE. See also [dm_examine_constraints()] for checking all constraints in a dm.

dm_add_pk(flights_dm, airports, tzone, check = TRUE)
#> Error: (`tzone`) not a unique key of `airports`.

Foreign keys

Useful functions for managing foreign key relations include:

  1. dm_add_fk()
  2. dm_get_all_fks()
  3. dm_rm_fk()
  4. dm_enum_fk_candidates()

Now it gets (even more) interesting: we want to define relations between different tables. With the dm_add_fk() function you can define which column of which table points to another table’s column.

This is done by choosing a foreign key from one table that will point to a primary key of another table. The primary key of the referred table must be set with dm_add_pk(). dm_add_fk() will find the primary key column of the referenced table by itself and make the indicated column of the child table point to it.

flights_dm_with_key %>% dm_add_fk(flights, origin, airports)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found

This will throw an error:

flights_dm %>% dm_add_fk(flights, origin, airports)
#> Error: (`origin`) is alreay a foreign key of table `flights` into table `airports`.

Let’s create a dm object with a foreign key relation to work with later on:

flights_dm_with_fk <- dm_add_fk(flights_dm_with_key, flights, origin, airports)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found

What if we tried to add another foreign key relation from flights to airports to the object? Column dest might work, since it also contains airport codes:

flights_dm_with_fk %>% dm_add_fk(flights, dest, airports, check = TRUE)
#> Error in is_dm(dm): object 'flights_dm_with_fk' not found

Checks are opt-in and executed only if check = TRUE. You can still add a foreign key with the default check = FALSE. See also [dm_examine_constraints()] for checking all constraints in a dm.

Get an overview of all foreign key relations withdm_get_all_fks():

dm_get_all_fks(dm_nycflights13(cycle = TRUE))
#> # A tibble: 5 x 4
#>   child_table child_fk_cols     parent_table parent_key_cols  
#>   <chr>       <keys>            <chr>        <keys>           
#> 1 flights     carrier           airlines     carrier          
#> 2 flights     origin            airports     faa              
#> 3 flights     dest              airports     faa              
#> 4 flights     tailnum           planes       tailnum          
#> 5 flights     origin, time_hour weather      origin, time_hour

Remove foreign key relations with dm_rm_fk() (parameter columns = NULL means that all relations will be removed, with a message):

flights_dm_with_fk %>%
  dm_rm_fk(table = flights, column = dest, ref_table = airports) %>%
  dm_get_fk(flights, airports)
#> Error in is_dm(dm): object 'flights_dm_with_fk' not found
flights_dm_with_fk %>%
  dm_rm_fk(flights, origin, airports) %>%
  dm_get_fk(flights, airports)
#> Error in is_dm(dm): object 'flights_dm_with_fk' not found
flights_dm_with_fk %>%
  dm_rm_fk(flights, columns = NULL, airports) %>%
  dm_get_fk(flights, airports)
#> Error in is_dm(dm): object 'flights_dm_with_fk' not found

Since the primary keys are defined in the dm object, you do not usually need to provide the referenced column name of ref_table.

Another function for getting to know your data better (cf. dm_enum_pk_candidates() in “Primary keys of dm objects”) is dm_enum_fk_candidates(). Use it to get an overview over foreign key candidates that point from one table to another:

dm_enum_fk_candidates(flights_dm_with_key, weather, airports)
#> Error in is_dm(dm): object 'flights_dm_with_key' not found