maurolepore / adm

Automatic Data Models

adm

The goal of adm is to provide automated tools to work with data models. It’s enabled by a lightweight convention on top of the dm package. Your data model must follow only two rules:

1. For each table, the primary key has the same name as the table with the prefix _id. For example the table x has a primary key named x_id.

2. If an _id column exists in a table which name is different than the prefix of that _id column, then that column is a foreign key linking to another table. For example, in the table x_id, the column y_id is a foreign key linking to the table y.

Installation

You can install the development version of adm from GitHub with:

# install.packages("devtools")
devtools::install_github("maurolepore/adm")

Example

library(dm, warn.conflicts = FALSE)
library(adm)

Consider this data model. The table and column names follow the three rules above, but the data model has no key yet.

adm_example <- adm_demo()
# Removing keys to show how to add them
dm <- dm(!!!as.list(adm_example))
dm
#> ── Metadata ────────────────────────────────────────────────────────────────────
#> Tables: `a`, `a_b`, `b`, `b_c`, `c`, … (7 total)
#> Columns: 17
#> Primary keys: 0
#> Foreign keys: 0

• adm::add_keys() adds primary and foreign keys.

adm <- dm %>% add_keys()
adm
#> ── Metadata ────────────────────────────────────────────────────────────────────
#> Tables: `a`, `a_b`, `b`, `b_c`, `c`, … (7 total)
#> Columns: 17
#> Primary keys: 7
#> Foreign keys: 5

adm %>% dm_draw()

• adm::flatten_linked_tables() flattens linked tables in a single one. Unlinked tables are excluded because there is no way to join them.

adm %>% flatten_linked_tables()
#> Warning: Unlinked tables: e
#> Joining with `by = join_by(b_id)`
#> Joining with `by = join_by(c_id)`
#> Joining with `by = join_by(a_id)`
#> Joining with `by = join_by(b_id)`
#> Joining with `by = join_by(d_id)`
#> # A tibble: 1 × 10
#>    a_id     a a_b_id  b_id     b b_c_id  c_id     c  d_id     d
#>   <dbl> <dbl>  <dbl> <dbl> <dbl>  <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1     1     1      1     1     1      1     1     1     1     1

Without adm

You can do the same directly with the dm package, but with complex data models it’s hard work.

• Add primary and foreign keys.

dm2 <- dm %>%
  dm_add_pk(a, a_id) %>%
  dm_add_pk(a_b, a_b_id) %>%
  dm_add_pk(b, b_id) %>%
  dm_add_pk(b_c, b_c_id) %>%
  dm_add_pk(c, c_id) %>%
  dm_add_pk(d, d_id) %>%
  dm_add_pk(e, e_id) %>%
  dm_add_fk(a_b, a_id, a) %>%
  dm_add_fk(a_b, b_id, b) %>%
  dm_add_fk(b_c, b_id, b) %>%
  dm_add_fk(b_c, c_id, c) %>%
  dm_add_fk(c, d_id, d)
dm2
#> ── Metadata ────────────────────────────────────────────────────────────────────
#> Tables: `a`, `a_b`, `b`, `b_c`, `c`, … (7 total)
#> Columns: 17
#> Primary keys: 7
#> Foreign keys: 5

dm2 %>% dm_draw()

• Flatten linked tables into a single one.

flat1 <- dm2 %>% dm_flatten_to_tbl(.start = a_b)
flat2 <- dm2 %>% dm_flatten_to_tbl(.start = b_c, .recursive = TRUE)
left_join(flat1, flat2)
#> Joining with `by = join_by(b_id, b)`
#> # A tibble: 1 × 10
#>   a_b_id  a_id  b_id     a     b b_c_id  c_id     c  d_id     d
#>    <dbl> <dbl> <dbl> <dbl> <dbl>  <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1      1     1     1     1     1      1     1     1     1     1