replyr: Get a Grip on Big Data in R
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.
replyr is an R package that contains extensions, adaptions, and work-arounds to make remote R dplyr data sources (including big data systems such as Spark) behave more like local data. This allows the analyst to more easily develop and debug procedures that simultaneously work on a variety of data services (in-memory data.frame, SQLite, PostgreSQL, and Spark2 currently being the primary supported platforms).
Example
Suppose we had a large data set hosted on a Spark cluster that we wished to work with using dplyr and sparklyr (for this article we will simulate such using data loaded into Spark from the nycflights13 package).
We will work a trivial example: taking a quick peek at your data. The analyst should always be able to and willing to look at the data.
It is easy to look at the top of the data, or any specific set of rows of the data.
Either through print() (which is much safter with tbl_df derived classes, than with base data.frame).
print(flts)
## Source: query [3.368e+05 x 19] ## Database: spark connection master=local[4] app=sparklyr local=TRUE ## ## year month day dep_time sched_dep_time dep_delay arr_time ## <int> <int> <int> <int> <int> <dbl> <int> ## 1 2013 1 1 517 515 2 830 ## 2 2013 1 1 533 529 4 850 ## 3 2013 1 1 542 540 2 923 ## 4 2013 1 1 544 545 -1 1004 ## 5 2013 1 1 554 600 -6 812 ## 6 2013 1 1 554 558 -4 740 ## 7 2013 1 1 555 600 -5 913 ## 8 2013 1 1 557 600 -3 709 ## 9 2013 1 1 557 600 -3 838 ## 10 2013 1 1 558 600 -2 753 ## # ... with 3.368e+05 more rows, and 12 more variables: ## # sched_arr_time <int>, arr_delay <dbl>, carrier <chr>, flight <int>, ## # tailnum <chr>, origin <chr>, dest <chr>, air_time <dbl>, ## # distance <dbl>, hour <dbl>, minute <dbl>, time_hour <dbl>
Or with dplyr::glimpse():
dplyr::glimpse(flts)
## Observations: 3.368e+05 ## Variables: 19 ## $ year <int> 2013, 2013, 2013, 2013, 2013, 2013, 2013, 2013,... ## $ month <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ day <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ dep_time <int> 517, 533, 542, 544, 554, 554, 555, 557, 557, 55... ## $ sched_dep_time <int> 515, 529, 540, 545, 600, 558, 600, 600, 600, 60... ## $ dep_delay <dbl> 2, 4, 2, -1, -6, -4, -5, -3, -3, -2, -2, -2, -2... ## $ arr_time <int> 830, 850, 923, 1004, 812, 740, 913, 709, 838, 7... ## $ sched_arr_time <int> 819, 830, 850, 1022, 837, 728, 854, 723, 846, 7... ## $ arr_delay <dbl> 11, 20, 33, -18, -25, 12, 19, -14, -8, 8, -2, -... ## $ carrier <chr> "UA", "UA", "AA", "B6", "DL", "UA", "B6", "EV",... ## $ flight <int> 1545, 1714, 1141, 725, 461, 1696, 507, 5708, 79... ## $ tailnum <chr> "N14228", "N24211", "N619AA", "N804JB", "N668DN... ## $ origin <chr> "EWR", "LGA", "JFK", "JFK", "LGA", "EWR", "EWR"... ## $ dest <chr> "IAH", "IAH", "MIA", "BQN", "ATL", "ORD", "FLL"... ## $ air_time <dbl> 227, 227, 160, 183, 116, 150, 158, 53, 140, 138... ## $ distance <dbl> 1400, 1416, 1089, 1576, 762, 719, 1065, 229, 94... ## $ hour <dbl> 5, 5, 5, 5, 6, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5,... ## $ minute <dbl> 15, 29, 40, 45, 0, 58, 0, 0, 0, 0, 0, 0, 0, 0, ... ## $ time_hour <dbl> 2013, 2013, 2013, 2013, 2013, 2013, 2013, 2013,...
What replyr adds to the task of “looking at the data” is a rough equivalent to base::summary(): a few per-column statistics.
# using dev version of replyr https://github.com/WinVector/replyr
replyr::replyr_summary(flts,
countUniqueNonNum= TRUE)
## column index class nrows nna nunique min max ## 1 year 1 integer 336776 0 NA 2013 2013 ## 2 month 2 integer 336776 0 NA 1 12 ## 3 day 3 integer 336776 0 NA 1 31 ## 4 dep_time 4 integer 336776 8255 NA 1 2400 ## 5 sched_dep_time 5 integer 336776 0 NA 106 2359 ## 6 dep_delay 6 numeric 336776 8255 NA -43 1301 ## 7 arr_time 7 integer 336776 8713 NA 1 2400 ## 8 sched_arr_time 8 integer 336776 0 NA 1 2359 ## 9 arr_delay 9 numeric 336776 9430 NA -86 1272 ## 10 carrier 10 character 336776 0 16 NA NA ## 11 flight 11 integer 336776 0 NA 1 8500 ## 12 tailnum 12 character 336776 0 4044 NA NA ## 13 origin 13 character 336776 0 3 NA NA ## 14 dest 14 character 336776 0 105 NA NA ## 15 air_time 15 numeric 336776 9430 NA 20 695 ## 16 distance 16 numeric 336776 0 NA 17 4983 ## 17 hour 17 numeric 336776 0 NA 1 23 ## 18 minute 18 numeric 336776 0 NA 0 59 ## 19 time_hour 19 numeric 336776 0 NA 2013 2013 ## mean sd lexmin lexmax ## 1 2013.000000 0.000000 <NA> <NA> ## 2 6.548510 3.414457 <NA> <NA> ## 3 15.710787 8.768607 <NA> <NA> ## 4 1349.109947 488.281791 <NA> <NA> ## 5 1344.254840 467.335756 <NA> <NA> ## 6 12.639070 40.210061 <NA> <NA> ## 7 1502.054999 533.264132 <NA> <NA> ## 8 1536.380220 497.457142 <NA> <NA> ## 9 6.895377 44.633292 <NA> <NA> ## 10 NA NA 9E YV ## 11 1971.923620 1632.471938 <NA> <NA> ## 12 NA NA <NA> <NA> ## 13 NA NA EWR LGA ## 14 NA NA ABQ XNA ## 15 150.686460 93.688305 <NA> <NA> ## 16 1039.912604 733.233033 <NA> <NA> ## 17 13.180247 4.661316 <NA> <NA> ## 18 26.230100 19.300846 <NA> <NA> ## 19 2013.000000 0.000000 <NA> <NA>
As we see, replyr summary returns data in a data frame, and can deal with multiple column types.
Note: the above summary has problems with NA in character columns with Spark, and thus is mis-reporting the NA count in the tailum column. We are working on the issue. That is also one of the advantages of taking your work-arounds from a package: when they do improve you can easily incorporate bring the improvements into your own work by a mere package update.
We could also use dplyr::summarize_each for the task, but it has the minor downside of returning the data in a wide form.
# currently throws if tailnum left in column list
vars <- setdiff(colnames(flts), 'tailnum')
flts %>% summarize_each(funs(min, max, mean, sd),
one_of(vars))
## Source: query [1 x 72] ## Database: spark connection master=local[4] app=sparklyr local=TRUE ## ## year_min month_min day_min dep_time_min sched_dep_time_min dep_delay_min ## <int> <int> <int> <int> <int> <dbl> ## 1 2013 1 1 1 106 -43 ## # ... with 66 more variables: arr_time_min <int>, ## # sched_arr_time_min <int>, arr_delay_min <dbl>, carrier_min <chr>, ## # flight_min <int>, origin_min <chr>, dest_min <chr>, ## # air_time_min <dbl>, distance_min <dbl>, hour_min <dbl>, ## # minute_min <dbl>, time_hour_min <dbl>, year_max <int>, ## # month_max <int>, day_max <int>, dep_time_max <int>, ## # sched_dep_time_max <int>, dep_delay_max <dbl>, arr_time_max <int>, ## # sched_arr_time_max <int>, arr_delay_max <dbl>, carrier_max <chr>, ## # flight_max <int>, origin_max <chr>, dest_max <chr>, ## # air_time_max <dbl>, distance_max <dbl>, hour_max <dbl>, ## # minute_max <dbl>, time_hour_max <dbl>, year_mean <dbl>, ## # month_mean <dbl>, day_mean <dbl>, dep_time_mean <dbl>, ## # sched_dep_time_mean <dbl>, dep_delay_mean <dbl>, arr_time_mean <dbl>, ## # sched_arr_time_mean <dbl>, arr_delay_mean <dbl>, carrier_mean <dbl>, ## # flight_mean <dbl>, origin_mean <dbl>, dest_mean <dbl>, ## # air_time_mean <dbl>, distance_mean <dbl>, hour_mean <dbl>, ## # minute_mean <dbl>, time_hour_mean <dbl>, year_sd <dbl>, ## # month_sd <dbl>, day_sd <dbl>, dep_time_sd <dbl>, ## # sched_dep_time_sd <dbl>, dep_delay_sd <dbl>, arr_time_sd <dbl>, ## # sched_arr_time_sd <dbl>, arr_delay_sd <dbl>, carrier_sd <dbl>, ## # flight_sd <dbl>, origin_sd <dbl>, dest_sd <dbl>, air_time_sd <dbl>, ## # distance_sd <dbl>, hour_sd <dbl>, minute_sd <dbl>, time_hour_sd <dbl>
Special code for remote data is needed as none of the obvious “one liner” candidates (base::summary(), or broom:glance()) are not currently (as of March 4, 2017) intended to work with remote data sources.
summary(flts)
## Length Class Mode ## src 1 src_spark list ## ops 3 op_base_remote list
str(flts)
## List of 2 ## $ src:List of 1 ## ..$ con:List of 11 ## .. ..$ master : chr "local[4]" ## .. ..$ method : chr "shell" ## .. ..$ app_name : chr "sparklyr" ## ...
packageVersion('broom')
## [1] '0.4.2'
broom::glance(flts)
## Error: glance doesn't know how to deal with data of class tbl_sparktbl_sqltbl_lazytbl
The source for the examples can be found here.
Conclusion
replyr_summary is not the only service replyr supplies, replyr includes many more adaptions including my own version of case-completion.
Roughly replyr is where I collect my adaptions so they don’t infest application code. replyr a way you can use heavy-duty big-data machinery, while keeping you fingers out of the gears.
R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.