dbplot

Leverages dplyr to process the calculations of a plot inside a database. This package provides helper functions that abstract the work at three levels:

  1. Functions that ouput a ggplot2 object
  2. Functions that outputs a data.frame object with the calculations
  3. Creates the formula needed to calculate bins for a Histogram or a Raster plot

Installation

# You can install the released version from CRAN
install.packages("dbplot")

# Or the the development version from GitHub:
install.packages("devtools")
devtools::install_github("edgararuiz/dbplot")

Connecting to a data source

Example

In addition to database connections, the functions work with sparklyr. A Spark DataFrame will be used for the examples in this README.

library(sparklyr)
sc <- spark_connect(master = "local", version = "2.1.0")
spark_flights <- copy_to(sc, nycflights13::flights, "flights")

ggplot

Histogram

By default dbplot_histogram() creates a 30 bin histogram

library(ggplot2)

spark_flights %>% 
  dbplot_histogram(sched_dep_time)

Use binwidth to fix the bin size

spark_flights %>% 
  dbplot_histogram(sched_dep_time, binwidth = 200)

Because it outputs a ggplot2 object, more customization can be done

spark_flights %>% 
  dbplot_histogram(sched_dep_time, binwidth = 300) +
  labs(title = "Flights - Scheduled Departure Time") +
  theme_bw()

Raster

To visualize two continuous variables, we typically resort to a Scatter plot. However, this may not be practical when visualizing millions or billions of dots representing the intersections of the two variables. A Raster plot may be a better option, because it concentrates the intersections into squares that are easier to parse visually.

A Raster plot basically does the same as a Histogram. It takes two continuous variables and creates discrete 2-dimensional bins represented as squares in the plot. It then determines either the number of rows inside each square or processes some aggregation, like an average.

  • If no fill argument is passed, the default calculation will be count, n()
spark_flights %>%
  filter(!is.na(arr_delay)) %>%
  dbplot_raster(arr_delay, dep_delay) 

  • Pass an aggregation formula that can run inside the database
spark_flights %>%
  filter(!is.na(arr_delay)) %>%
  dbplot_raster(arr_delay, dep_delay, mean(distance, na.rm = TRUE)) 

  • Increase or decrease for more, or less, definition. The resolution argument controls that, it defaults to 100
spark_flights %>%
  filter(!is.na(arr_delay)) %>%
  dbplot_raster(arr_delay, dep_delay, mean(distance, na.rm = TRUE), resolution = 500)

Bar Plot

  • dbplot_bar() defaults to a tally() of each value in a discrete variable
spark_flights %>%
  dbplot_bar(origin)

  • Pass a formula that will be operated for each value in the discrete variable
spark_flights %>%
  dbplot_bar(origin, mean(dep_delay))
## Warning: Missing values are always removed in SQL.
## Use `AVG(x, na.rm = TRUE)` to silence this warning

Line plot

  • dbplot_line() defaults to a tally() of each value in a discrete variable
spark_flights %>%
  dbplot_line(month)

  • Pass a formula that will be operated for each value in the discrete variable
spark_flights %>%
  dbplot_line(month, mean(dep_delay))
## Warning: Missing values are always removed in SQL.
## Use `AVG(x, na.rm = TRUE)` to silence this warning

Boxplot

  • It expects a discrete variable to group by, and a continuous variable to calculate the percentiles and IQR. It doesn’t calculate outliers. Currently, this feature works with sparklyr and Hive connections.
spark_flights %>%
  dbplot_boxplot(origin, dep_delay)

Calculation functions

If a more customized plot is needed, the data the underpins the plots can also be accessed:

  1. db_compute_bins() - Returns a data frame with the bins and count per bin
  2. db_compute_count() - Returns a data frame with the count per discrete value
  3. db_compute_raster() - Returns a data frame with the results per x/y intersection
  4. db_compute_boxplot() - Returns a data frame with boxplot calculations
spark_flights %>%
  db_compute_bins(arr_delay) 
## # A tibble: 28 x 2
##    arr_delay     count
##        <dbl>     <dbl>
##  1      4.53  79784   
##  2   - 40.7  207999   
##  3     95.1    7890   
##  4     49.8   19063   
##  5    819         8.00
##  6    140      3746   
##  7    321       232   
##  8    231       921   
##  9   - 86.0    5325   
## 10    186      1742   
## # ... with 18 more rows

The data can be piped to a plot

spark_flights %>%
  filter(arr_delay < 100 , arr_delay > -50) %>%
  db_compute_bins(arr_delay) %>%
  ggplot() +
  geom_col(aes(arr_delay, count, fill = count))

db_bin()

Uses ‘rlang’ to build the formula needed to create the bins of a numeric variable in an un-evaluated fashion. This way, the formula can be then passed inside a dplyr verb.

db_bin(var)
## (((max(var, na.rm = TRUE) - min(var, na.rm = TRUE))/(30)) * ifelse((as.integer(floor(((var) - 
##     min(var, na.rm = TRUE))/((max(var, na.rm = TRUE) - min(var, 
##     na.rm = TRUE))/(30))))) == (30), (as.integer(floor(((var) - 
##     min(var, na.rm = TRUE))/((max(var, na.rm = TRUE) - min(var, 
##     na.rm = TRUE))/(30))))) - 1, (as.integer(floor(((var) - min(var, 
##     na.rm = TRUE))/((max(var, na.rm = TRUE) - min(var, na.rm = TRUE))/(30))))))) + 
##     min(var, na.rm = TRUE)
spark_flights %>%
  group_by(x = !! db_bin(arr_delay)) %>%
  tally()
## # Source:   lazy query [?? x 2]
## # Database: spark_connection
##          x         n
##      <dbl>     <dbl>
##  1    4.53  79784   
##  2 - 40.7  207999   
##  3   95.1    7890   
##  4   49.8   19063   
##  5  819         8.00
##  6  140      3746   
##  7  321       232   
##  8  231       921   
##  9 - 86.0    5325   
## 10  186      1742   
## # ... with more rows
spark_flights %>%
  filter(!is.na(arr_delay)) %>%
  group_by(x = !! db_bin(arr_delay)) %>%
  tally()%>%
  collect %>%
  ggplot() +
  geom_col(aes(x, n))