dbplyr: SQL ผ่านไวยากรณ์ dplyr

dbplyr คืออะไร

# install.packages(c('dbplyr', 'DBI', 'RSQLite'))
library(DBI)
library(dbplyr)
library(dplyr)

# dbplyr sits between dplyr and your database:
# Your dplyr code -> dbplyr -> SQL -> Database -> result

# Supported backends: PostgreSQL, MySQL, SQLite,
#                     SQL Server, BigQuery, Snowflake, ...

cat('dbplyr translates dplyr to SQL')

การเชื่อมต่อกับฐานข้อมูล

library(DBI)
library(dplyr)
library(dbplyr)

# SQLite example (no server needed — great for demos)
con <- dbConnect(RSQLite::SQLite(), ':memory:')

# Write a test table to the in-memory database
copy_to(con, nycflights13::flights, 'flights',
        temporary = FALSE, overwrite = TRUE)

# PostgreSQL example (real server):
# con <- dbConnect(
#   RPostgres::Postgres(),
#   host     = 'db.example.com',
#   dbname   = 'analytics',
#   user     = Sys.getenv('DB_USER'),
#   password = Sys.getenv('DB_PASS')
# )

cat('Connected to database')

tbl() — อ้างอิงตารางฐานข้อมูล

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
mtcars_df <- mtcars
dbWriteTable(con, 'cars', mtcars_df)

# Create a lazy table reference
cars_tbl <- tbl(con, 'cars')

# Printing shows top rows and indicates it is a database source
print(cars_tbl)
# Source: table<cars> [?? x 11]
# Database: sqlite 3.x [:memory:]

cat('tbl() = lazy reference, no data fetched yet')

การใช้ filter() กับตารางฐานข้อมูล

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
cars_tbl <- tbl(con, 'cars')

# Filter in the database (generates WHERE clause)
high_mpg <- cars_tbl |>
  filter(mpg > 25, cyl == 4)

# Still lazy! No data fetched yet.
cat('Class:', class(high_mpg)[1], '\n')

# Collect to pull data into R:
result <- collect(high_mpg)
cat('Rows matching filter:', nrow(result))

การใช้ select() และ mutate()

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
cars_tbl <- tbl(con, 'cars')

# Select specific columns + add a computed column
result <- cars_tbl |>
  filter(am == 1) |>              # manual transmission
  select(mpg, cyl, hp, wt) |>    # pick columns
  mutate(wt_kg = wt * 453.592)   # add computed column

# Pull into R
df <- collect(result)
cat('Columns:', names(df), '\n')
cat('Rows:', nrow(df))

group_by() และ summarise() — การรวมข้อมูล

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
cars_tbl <- tbl(con, 'cars')

# Aggregate in the database
summary_tbl <- cars_tbl |>
  group_by(cyl) |>
  summarise(
    avg_mpg  = mean(mpg, na.rm = TRUE),
    max_hp   = max(hp),
    n_models = n()
  )

result <- collect(summary_tbl)
print(result)

show_query() — ตรวจสอบ SQL ที่สร้างขึ้น

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
cars_tbl <- tbl(con, 'cars')

# Build a query
q <- cars_tbl |>
  filter(mpg > 20) |>
  group_by(cyl) |>
  summarise(avg_mpg = mean(mpg, na.rm = TRUE))

# See the SQL dbplyr generated:
show_query(q)
# <SQL>
# SELECT cyl, AVG(mpg) AS avg_mpg
# FROM cars
# WHERE mpg > 20.0
# GROUP BY cyl

collect() — ดึงข้อมูลเข้าสู่ R

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
cars_tbl <- tbl(con, 'cars')

# Build up a lazy query chain
lazy_q <- cars_tbl |>
  filter(hp > 100) |>
  select(mpg, hp, cyl) |>
  arrange(desc(hp))

# Nothing fetched yet
cat('Is lazy?', inherits(lazy_q, 'tbl_sql'), '\n')

# NOW pull data into R
local_df <- collect(lazy_q)
cat('Class after collect:', class(local_df)[1], '\n')
cat('Rows:', nrow(local_df))

copy_to() — ส่ง data frame ภายในเครื่องไปยังฐานข้อมูล

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')

# Push a local data frame into the database
local_df <- data.frame(
  cyl       = c(4, 6, 8),
  category  = c('efficient', 'balanced', 'powerful')
)

# copy_to creates a temporary table in the DB
copy_to(con, local_df, name = 'cyl_labels', temporary = TRUE)

# Now reference it with tbl()
labels_tbl <- tbl(con, 'cyl_labels')
cat('Rows in DB table:', collect(labels_tbl) |> nrow())

การเชื่อมตารางฐานข้อมูล

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
copy_to(con, data.frame(cyl = c(4,6,8),
                         label = c('four','six','eight')),
        'cyl_ref', temporary = TRUE)

cars_tbl  <- tbl(con, 'cars')
labels_tbl <- tbl(con, 'cyl_ref')

# Join in the database
joined <- left_join(cars_tbl, labels_tbl, by = 'cyl') |>
  select(mpg, cyl, label, hp)

show_query(joined)   # See the SQL JOIN
result <- collect(joined)
cat('Joined rows:', nrow(result))

กรณีที่ไม่ควรใช้ dbplyr

library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'cars', mtcars)
cars_tbl <- tbl(con, 'cars')

# Do as much as possible in the database:
prepped <- cars_tbl |>
  filter(mpg > 15) |>
  select(mpg, hp, cyl) |>
  collect()               # <- pull only what you need

# Now apply R-only operations locally:
cor_result <- cor(prepped$mpg, prepped$hp)
cat('Correlation mpg~hp:', round(cor_result, 3))

# Rule: filter and aggregate in DB, model in R

ตรวจสอบความเข้าใจอย่างรวดเร็ว

dbplyr — ประเด็นสำคัญ

# Complete dbplyr workflow example:
library(DBI)
library(dplyr)
library(dbplyr)

con <- dbConnect(RSQLite::SQLite(), ':memory:')
dbWriteTable(con, 'sales', data.frame(
  region  = c('North','South','North','East','South'),
  revenue = c(100, 200, 150, 300, 250),
  year    = c(2023, 2023, 2024, 2024, 2024)
))

tbl(con, 'sales') |>
  filter(year == 2024) |>
  group_by(region) |>
  summarise(total = sum(revenue, na.rm = TRUE)) |>
  arrange(desc(total)) |>
  collect() |>
  print()