lisjoin
A small package with list-tibble join operations.
Motivation
There no native strictly key-pair data structure to R. There is, however, a sufficently good first-approach which is a named list.
{
ids <- paste0(sample(letters, size = 1000, replace = TRUE),
sample(1:30, 4, replace = TRUE))
named_list <- purrr::map(1:1000, function(.x) rnorm(10))
names(named_list) <- ids
head(named_list)
}## $p28
## [1] -1.46777476 0.74539593 -0.04664338 0.10781217 1.46081849 1.12868098
## [7] -0.92717928 -0.14309675 -0.11820866 -0.43120743
##
## $s30
## [1] -1.31831735 1.99491831 -0.03777807 -1.25668048 0.09732199 -0.65921933
## [7] 0.89431084 -0.36112306 -1.11437703 1.61232659
##
## $v12
## [1] -1.3033179675 1.1507490652 0.0002327135 -0.4352504670 -0.0690152133
## [6] 2.0782264295 2.0316954216 -1.0194318330 -1.2564915962 0.4674051620
##
## $f15
## [1] -2.52414903 -2.04408138 0.16168534 0.73952108 -0.98379734 -0.24529745
## [7] 0.39888376 -1.97808891 0.29995678 0.01493763
##
## $y28
## [1] -2.18879872 1.09157658 -0.04494610 -0.21365753 -1.61516204 0.92281453
## [7] -0.00809693 -0.78239806 -0.37681752 1.10060229
##
## $n30
## [1] -0.9895956 0.5019188 0.6067216 -0.7643856 1.1582335 0.2849196
## [7] -0.6200820 0.1745718 -0.3898789 -0.5189159
And this approach can get you far-ish.
library(tidyverse)## ── Attaching packages ────────────────────────────────────────────────────────────────────── tidyverse 1.3.0 ──
## ✓ ggplot2 3.3.2 ✓ purrr 0.3.4
## ✓ tibble 3.0.3 ✓ dplyr 1.0.2
## ✓ tidyr 1.1.2 ✓ stringr 1.4.0
## ✓ readr 1.3.1 ✓ forcats 0.5.0
## ── Conflicts ───────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()
pluck(named_list, sample(ids, 1)) ## [1] 0.593759344 0.009909793 -0.743166196 0.221449062 0.530201613
## [6] 0.078539020 0.606547272 0.335228779 -0.270646825 -1.218436378
# update specific values
ids_to_update <- sample(ids, 5)
modify_at(named_list,
ids_to_update,
function(x) 2*x - 3) ->
updated_list
sample(named_list, 200) %>% # recover ids given condition
keep(~ mean(.x) < -.5) %>%
names()## [1] "w28" "y12" "t15" "k30" "e15" "b12" "u28" "i15" "l28" "s28" "c30" "j28"
## [13] "g15" "f28" "g30"
And by that I mean this far.
(tibble(key = sample(ids, 100),
val = runif(100)) ->
keyval_tibble)## # A tibble: 100 x 2
## key val
## <chr> <dbl>
## 1 m15 0.291
## 2 j28 0.686
## 3 f28 0.930
## 4 e30 0.889
## 5 o30 0.977
## 6 g28 0.914
## 7 t28 0.790
## 8 q28 0.0524
## 9 y15 0.644
## 10 t15 0.764
## # … with 90 more rows
left_join(keyval_tibble)## Error in auto_copy(x, y, copy = copy): argument "y" is missing, with no default
left_join(keyval_tibble, named_list, by = character(), copy = TRUE)## # A tibble: 1,000 x 1,002
## key val p28 s30 v12 f15 y28 n30 c12 c15
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 m15 0.291 -1.47 -1.32 -1.30e+0 -2.52 -2.19 -0.990 -1.62 0.353
## 2 m15 0.291 0.745 1.99 1.15e+0 -2.04 1.09 0.502 -0.438 0.571
## 3 m15 0.291 -0.0466 -0.0378 2.33e-4 0.162 -0.0449 0.607 -0.474 1.70
## 4 m15 0.291 0.108 -1.26 -4.35e-1 0.740 -0.214 -0.764 -0.381 1.24
## 5 m15 0.291 1.46 0.0973 -6.90e-2 -0.984 -1.62 1.16 -1.48 -0.319
## 6 m15 0.291 1.13 -0.659 2.08e+0 -0.245 0.923 0.285 -0.984 0.431
## 7 m15 0.291 -0.927 0.894 2.03e+0 0.399 -0.00810 -0.620 -0.532 0.968
## 8 m15 0.291 -0.143 -0.361 -1.02e+0 -1.98 -0.782 0.175 -0.331 0.378
## 9 m15 0.291 -0.118 -1.11 -1.26e+0 0.300 -0.377 -0.390 1.37 -0.504
## 10 m15 0.291 -0.431 1.61 4.67e-1 0.0149 1.10 -0.519 -1.54 -1.20
## # … with 990 more rows, and 992 more variables: h28 <dbl>, i30 <dbl>,
## # t12 <dbl>, h15 <dbl>, c28 <dbl>, w30 <dbl>, w12 <dbl>, a15 <dbl>,
## # x28 <dbl>, k30 <dbl>, w12.1 <dbl>, g15 <dbl>, g28 <dbl>, e30 <dbl>,
## # n12 <dbl>, n15 <dbl>, o28 <dbl>, a30 <dbl>, o12 <dbl>, z15 <dbl>,
## # m28 <dbl>, b30 <dbl>, c12.1 <dbl>, o15 <dbl>, w28 <dbl>, d30 <dbl>,
## # w12.2 <dbl>, m15 <dbl>, b28 <dbl>, j30 <dbl>, m12 <dbl>, o15.1 <dbl>,
## # s28 <dbl>, q30 <dbl>, t12.1 <dbl>, t15 <dbl>, c28.1 <dbl>, d30.1 <dbl>,
## # r12 <dbl>, u15 <dbl>, b28.1 <dbl>, v30 <dbl>, a12 <dbl>, a15.1 <dbl>,
## # h28.1 <dbl>, f30 <dbl>, m12.1 <dbl>, l15 <dbl>, g28.1 <dbl>, q30.1 <dbl>,
## # p12 <dbl>, i15 <dbl>, b28.2 <dbl>, g30 <dbl>, x12 <dbl>, f15.1 <dbl>,
## # h28.2 <dbl>, e30.1 <dbl>, c12.2 <dbl>, p15 <dbl>, x28.1 <dbl>, e30.2 <dbl>,
## # x12.1 <dbl>, n15.1 <dbl>, e28 <dbl>, y30 <dbl>, a12.1 <dbl>, x15 <dbl>,
## # q28 <dbl>, x30 <dbl>, l12 <dbl>, g15.1 <dbl>, f28 <dbl>, w30.1 <dbl>,
## # k12 <dbl>, t15.1 <dbl>, e28.1 <dbl>, u30 <dbl>, b12 <dbl>, z15.1 <dbl>,
## # a28 <dbl>, z30 <dbl>, b12.1 <dbl>, p15.1 <dbl>, q28.1 <dbl>, k30.1 <dbl>,
## # h12 <dbl>, h15.1 <dbl>, j28 <dbl>, p30 <dbl>, z12 <dbl>, u15.1 <dbl>,
## # u28 <dbl>, y30.1 <dbl>, c12.3 <dbl>, d15 <dbl>, f28.1 <dbl>, v30.1 <dbl>,
## # w12.3 <dbl>, q15 <dbl>, …
lisjoin, as of now, provides a lazy prototypical approach:
library(lisjoin)
lisjoin(keyval_tibble, named_list, .key = key)## # A tibble: 1,013 x 3
## key val list_val
## <chr> <dbl> <list>
## 1 m15 0.291 <dbl [10]>
## 2 m15 0.291 <dbl [10]>
## 3 m15 0.291 <dbl [10]>
## 4 m15 0.291 <dbl [10]>
## 5 m15 0.291 <dbl [10]>
## 6 m15 0.291 <dbl [10]>
## 7 m15 0.291 <dbl [10]>
## 8 m15 0.291 <dbl [10]>
## 9 j28 0.686 <dbl [10]>
## 10 j28 0.686 <dbl [10]>
## # … with 1,003 more rows
Right now lisjoin supports key-guessing, type stability,
left/right/inner/full joins. For example, since we know the output is
going to be a double-precision number:
library(lisjoin)
map(named_list,
~ reduce(.x, sum)) %>%
lisjoin(keyval_tibble,
.,
.key = key,
type = 'dbl')## # A tibble: 1,013 x 3
## key val list_val
## <chr> <dbl> <dbl>
## 1 m15 0.291 -4.24
## 2 m15 0.291 -4.24
## 3 m15 0.291 -4.24
## 4 m15 0.291 -4.24
## 5 m15 0.291 -4.24
## 6 m15 0.291 -4.24
## 7 m15 0.291 -4.24
## 8 m15 0.291 -4.24
## 9 j28 0.686 0.376
## 10 j28 0.686 0.376
## # … with 1,003 more rows
Long term goals
This package is inspired by clojure’s map structure, purrr’s filosophy and design. In the long term I see it as a tool allowing tidy data workflows on lists.