Collapsing a 4-column data frame of real data from 500,000 rows to 91,000 by pasting and counting row values. Execution on a 1.8GHz Intel Core i5 shows that dplyr is 1.5 times quicker than base R.
See this Gist for a simpler test over twice more rows and roughly as many groups. In both tests, dplyr is as concise as plyr, as fast as data.table, and clearly more readable than base R.
| # grouping variable used for aggregation | |
| length(unique(sponsorships$uid)) | |
| [1] 91505 | |
| # used for %in% search during aggreation | |
| au = unique(sponsors$url) | |
| length(au) | |
| [1] 1499 | |
| # base | |
| assemble = function(sponsorships) { | |
| a = aggregate(name ~ uid, paste0, collapse = ",", data = subset(sponsorships, status == "author")) | |
| b = aggregate(name ~ uid, paste0, collapse = ",", data = subset(sponsorships, status == "cosponsor")) | |
| b = merge(a, b, by = "uid", all.x = TRUE) | |
| a = aggregate(url ~ uid, function(x) { sum(is.na(x)) }, data = sponsorships) | |
| b = merge(a, b, by = "uid", all.x = TRUE) | |
| a = aggregate(url ~ uid, function(x) { x %in% au }, data = subset(sponsorships, !is.na(url))) | |
| b = merge(b, a, by = "uid", all.x = TRUE) | |
| names(b)[-1] = c("n_missing", "auteurs", "cosignataires", "n_known") | |
| return(b[, c("uid", "auteurs", "cosignataires", "n_missing", "n_known")]) | |
| } | |
| system.time(assemble(sponsorships)) | |
| user system elapsed | |
| 12.397 1.218 13.828 | |
| # plyr | |
| library(plyr) | |
| au = unique(sponsors$url) | |
| system.time(auteurs <- ddply(subset(sponsorships, status %in% c("author", "cosponsor")), .(uid), | |
| summarise, | |
| auteurs = paste0(name[ grepl("author", status) ], collapse = ","), | |
| cosignataires = paste0(name[ grepl("cosponsor", status) ], collapse = ","), | |
| n_missing = sum(is.na(url)), | |
| n_known = sum(url %in% au) | |
| ) | |
| ) | |
| user system elapsed | |
| 166.870 94.133 264.323 | |
| # dplyr | |
| library(dplyr) | |
| system.time(sponsorships %.% | |
| filter(status %in% c("author", "cosponsor")) %.% | |
| group_by(uid) %.% | |
| summarise( | |
| auteurs = paste0(name[ status == "author" ], collapse = ","), | |
| cosignataires = paste0(name[ status == "cosponsor" ], collapse = ","), | |
| n_missing = sum(is.na(url)), | |
| n_known = sum(url %in% au) | |
| ) | |
| ) | |
| user system elapsed | |
| 7.191 0.941 8.164 | |
| # versions | |
| sessionInfo() | |
| R version 3.0.2 (2013-09-25) | |
| Platform: x86_64-apple-darwin10.8.0 (64-bit) | |
| locale: | |
| [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 | |
| attached base packages: | |
| [1] graphics grDevices datasets stats grid utils methods base | |
| other attached packages: | |
| [1] data.table_1.9.2 dplyr_0.1.2 XML_3.95-0.2 stringr_0.6.2 RPostgreSQL_0.4 DBI_0.2-7 network_1.9.0 | |
| [8] lubridate_1.3.3 knitr_1.5 ggmap_2.3 GGally_0.4.5 reshape_0.8.4 plyr_1.8.1 downloader_0.3 | |
| [15] devtools_1.4.1 ggplot2_0.9.3.1 | |
| loaded via a namespace (and not attached): | |
| [1] assertthat_0.1 colorspace_1.2-4 dichromat_2.0-0 digest_0.6.4 evaluate_0.5.1 formatR_0.10 | |
| [7] gtable_0.1.2 httr_0.2 labeling_0.2 mapproj_1.2-2 maps_2.3-6 MASS_7.3-29 | |
| [13] memoise_0.1 munsell_0.4.2 parallel_3.0.2 png_0.1-7 proto_0.3-10 RColorBrewer_1.0-5 | |
| [19] Rcpp_0.11.0 RCurl_1.95-4.1 reshape2_1.2.2 RgoogleMaps_1.2.0.5 rjson_0.2.13 RJSONIO_1.0-3 | |
| [25] scales_0.2.3 tools_3.0.2 whisker_0.3-2 |