Fix some lints and re-style

R/hamming_join_core.R

@@ -19,21 +19,21 @@ hamming_join <- function(a, b, mode, by, n_bands, band_width,
   stopifnot("'by' vectors must have length 1" = length(by_a) == 1)
   stopifnot("'by' vectors must have length 1" = length(by_b) == 1)
 
-  stopifnot("There should be no NA's in by_a" = !any(is.na(dplyr::pull(a, by_a))))
-  stopifnot("There should be no NA's in by_b" = !any(is.na(dplyr::pull(b, by_b))))
+  stopifnot("There should be no NA's in by_a" = !anyNA(a[[by_a]]))
+  stopifnot("There should be no NA's in by_b" = !anyNA(b[[by_b]]))
 
   # Clean strings that are matched on
   if (clean) {
-    a_col <- gsub("[[:punct:] ]", "", dplyr::pull(a, by_a))
-    b_col <- gsub("[[:punct:] ]", "", dplyr::pull(b, by_b))
+    a_col <- tolower(gsub("[[:punct:] ]", "", dplyr::pull(a, by_a)))
+    b_col <- tolower(gsub("[[:punct:] ]", "", dplyr::pull(b, by_b)))
   } else {
     a_col <- dplyr::pull(a, by_a)
     b_col <- dplyr::pull(b, by_b)
   }
 
   max_chars <- max(c(nchar(a_col), nchar(b_col)))
-
   thresh_prob <- hamming_probability(threshold, max_chars, n_bands, band_width)
+
   if (thresh_prob < .95) {
     str <- paste0(
       "A pair of records at the threshold (", threshold,
@@ -44,8 +44,6 @@ hamming_join <- function(a, b, mode, by, n_bands, band_width,
     warning(str)
   }
 
-
-
   match_table <- rust_hamming_join(
     a_col, b_col,
     band_width, n_bands, threshold,

R/hamming_logical_joins.R

@@ -73,100 +73,110 @@
 #'   band_width = 10,
 #' )
 hamming_inner_join <- function(a, b,
-                           by = NULL,
-                           n_bands = 100,
-                           band_width = 100,
-                           threshold = 2,
-                           progress = FALSE,
-                           clean = FALSE,
-                           similarity_column=NULL) {
-    hamming_join(a, b, mode = "inner",
-                 by = by,
-                 n_bands = n_bands,
-                 band_width = band_width,
-                 threshold =  threshold,
-                 progress = progress,
-                 similarity_column =  similarity_column,
-                 clean=clean)
+                               by = NULL,
+                               n_bands = 100,
+                               band_width = 100,
+                               threshold = 2,
+                               progress = FALSE,
+                               clean = FALSE,
+                               similarity_column = NULL) {
+  hamming_join(a, b,
+    mode = "inner",
+    by = by,
+    n_bands = n_bands,
+    band_width = band_width,
+    threshold = threshold,
+    progress = progress,
+    similarity_column = similarity_column,
+    clean = clean
+  )
 }
 
 #' @rdname hamming-joins
 #' @export
 hamming_anti_join <- function(a, b,
-                           by = NULL,
-                           n_bands = 100,
-                           band_width = 100,
-                           threshold = 2,
-                           progress = FALSE,
-                           clean = FALSE,
-                           similarity_column=NULL) {
-    hamming_join(a, b, mode = "anti",
-                 by = by,
-                 n_bands = n_bands,
-                 band_width = band_width,
-                 threshold =  threshold,
-                 progress = progress,
-                 similarity_column =  similarity_column,
-                 clean=clean)
+                              by = NULL,
+                              n_bands = 100,
+                              band_width = 100,
+                              threshold = 2,
+                              progress = FALSE,
+                              clean = FALSE,
+                              similarity_column = NULL) {
+  hamming_join(a, b,
+    mode = "anti",
+    by = by,
+    n_bands = n_bands,
+    band_width = band_width,
+    threshold = threshold,
+    progress = progress,
+    similarity_column = similarity_column,
+    clean = clean
+  )
 }
 
 #' @rdname hamming-joins
 #' @export
 hamming_left_join <- function(a, b,
-                           by = NULL,
-                           n_bands = 100,
-                           band_width = 100,
-                           threshold = 2,
-                           progress = FALSE,
-                           clean = FALSE,
-                           similarity_column=NULL) {
-    hamming_join(a, b, mode = "left",
-                 by = by,
-                 n_bands = n_bands,
-                 band_width = band_width,
-                 threshold =  threshold,
-                 progress = progress,
-                 similarity_column =  similarity_column,
-                 clean=clean)
+                              by = NULL,
+                              n_bands = 100,
+                              band_width = 100,
+                              threshold = 2,
+                              progress = FALSE,
+                              clean = FALSE,
+                              similarity_column = NULL) {
+  hamming_join(a, b,
+    mode = "left",
+    by = by,
+    n_bands = n_bands,
+    band_width = band_width,
+    threshold = threshold,
+    progress = progress,
+    similarity_column = similarity_column,
+    clean = clean
+  )
 }
 
 #' @rdname hamming-joins
 #' @export
 hamming_right_join <- function(a, b,
-                           by = NULL,
-                           n_bands = 100,
-                           band_width = 100,
-                           threshold = 2,
-                           progress = FALSE,
-                           clean = FALSE,
-                           similarity_column=NULL) {
-    hamming_join(a, b, mode = "right",
-                 by = by,
-                 n_bands = n_bands,
-                 band_width = band_width,
-                 threshold =  threshold,
-                 progress = progress,
-                 similarity_column =  similarity_column,
-                 clean=clean)
+                               by = NULL,
+                               n_bands = 100,
+                               band_width = 100,
+                               threshold = 2,
+                               progress = FALSE,
+                               clean = FALSE,
+                               similarity_column = NULL) {
+  hamming_join(a, b,
+    mode = "right",
+    by = by,
+    n_bands = n_bands,
+    band_width = band_width,
+    threshold = threshold,
+    progress = progress,
+    similarity_column = similarity_column,
+    clean = clean
+  )
 }
 
 
 #' @rdname hamming-joins
 #' @export
 hamming_full_join <- function(a, b,
-                           by = NULL,
-                           n_bands = 100,
-                           band_width = 100,
-                           threshold = 2,
-                           progress = FALSE,
-                           clean = FALSE,
-                           similarity_column=NULL) {
-    hamming_join(a, b, mode = "full",
-                 by = by,
-                 n_bands = n_bands,
-                 band_width = band_width,
-                 threshold =  threshold,
-                 progress = progress,
-                 similarity_column =  similarity_column,
-                 clean=clean)
+                              by = NULL,
+                              n_bands = 100,
+                              band_width = 100,
+                              threshold = 2,
+                              progress = FALSE,
+                              clean = FALSE,
+                              similarity_column = NULL) {
+  hamming_join(a, b,
+    mode = "full",
+    by = by,
+    n_bands = n_bands,
+    band_width = band_width,
+    threshold = threshold,
+    progress = progress,
+    similarity_column = similarity_column,
+    clean = clean
+  )
 }

R/jaccard_similarity.R

@@ -29,11 +29,13 @@ jaccard_similarity <- function(a, b, ngram_width = 2) {
 #' @return a vector of hamming similarities of the strings
 #'
 #' @examples
-#' hamming_distance(c("the quick brown fox","jumped over the lazy dog"),
-#'     c("the quck bron fx","jumped over hte lazy dog"))
+#' hamming_distance(
+#'   c("the quick brown fox", "jumped over the lazy dog"),
+#'   c("the quck bron fx", "jumped over hte lazy dog")
+#' )
 #'
 #' @export
-hamming_distance <- function(a, b, ngram_width=2) {
-    stopifnot(length(a) == length(b))
-    rust_hamming_distance(a, b)
+hamming_distance <- function(a, b, ngram_width = 2) {
+  stopifnot(length(a) == length(b))
+  rust_hamming_distance(a, b)
 }

R/lsh_properties.R

@@ -180,20 +180,20 @@ jaccard_hyper_grid_search <- function(s1 = .1, s2 = .7, p1 = .001, p2 = .999) {
 #'
 #' @export
 hamming_probability <- function(distance, input_length, n_bands, band_width) {
-    # probability that two strings with distance d have same value for randomly
-    # chosen bit
-    p_one_collision <- 1-(distance/input_length)
+  # probability that two strings with distance d have same value for randomly
+  # chosen bit
+  p_one_collision <- 1 - (distance / input_length)
 
-    # probability that two strings with distance d have same value for band_width
-    # randomly chosen bits
-    p_one_band <- p_one_collision^band_width
+  # probability that two strings with distance d have same value for band_width
+  # randomly chosen bits
+  p_one_band <- p_one_collision^band_width
 
-    # probability that two strings with distance d have same value for one of any
-    # n_bands hashes
-    # Pr[compared] = 1 - Pr[no hashes match]
-    #              = 1 - Pr[one hash does not match]^n_bands
+  # probability that two strings with distance d have same value for one of any
+  # n_bands hashes
+  # Pr[compared] = 1 - Pr[no hashes match]
+  #              = 1 - Pr[one hash does not match]^n_bands
 
-    p_compared <- 1-(1-p_one_band)^n_bands
+  p_compared <- 1 - (1 - p_one_band)^n_bands
 
-    return(p_compared)
+  return(p_compared)
 }

tests/testthat/test-hamming_dist.R

@@ -1,5 +1,4 @@
 test_that("hamming distance works", {
-
   require(babynames)
   require(stringdist)
 
@@ -9,12 +8,11 @@ test_that("hamming distance works", {
 
 
     a <- hamming_distance(nameys, shuff_nameys)
-    b <- stringdist(nameys, shuff_nameys,  method = "hamming")
+    b <- stringdist(nameys, shuff_nameys, method = "hamming")
 
-    expect_true(all((a == Inf) %in% (b==Inf)))
-    expect_true(all((b == Inf) %in% (a==Inf)))
+    expect_true(all((a == Inf) %in% (b == Inf)))
+    expect_true(all((b == Inf) %in% (a == Inf)))
 
     expect_true(all(abs(a[a != Inf] - b[b != Inf]) < .01))
-
   }
 })

tests/testthat/test-test_jaccard_sim.R

@@ -12,4 +12,3 @@ test_that("jaccard sim works", {
     expect_true(all(abs(a - b) < .01))
   }
 })
-

tests/testthat/test-test_logical_lsh_join.R

@@ -183,7 +183,7 @@ test_that("hamming_full_join works on tiny dataset", {
 
 test_that("hamming_left_join works on tiny dataset", {
   capture_messages(
-    test <- hamming_left_join(dataset_1, dataset_2, threshold = 3, band_width = 1,  n_bands = 300)
+    test <- hamming_left_join(dataset_1, dataset_2, threshold = 3, band_width = 1, n_bands = 300)
   )
 
   expect_true(all(test$id_1 == test$id_2, na.rm = T))
@@ -194,7 +194,7 @@ test_that("hamming_left_join works on tiny dataset", {
 
 test_that("hamming_right_join works on tiny dataset", {
   capture_messages(
-    test <- hamming_right_join(dataset_1, dataset_2, threshold = 3, band_width = 1,  n_bands = 300)
+    test <- hamming_right_join(dataset_1, dataset_2, threshold = 3, band_width = 1, n_bands = 300)
   )
 
 
@@ -207,7 +207,6 @@ test_that("hamming_right_join works on tiny dataset", {
 test_that("jaccard_inner_join gives same results as stringdist_inner_join", {
   for (i in 1:20) {
     capture_messages({
-
       zoomer_join_out <- hamming_inner_join(names_df, misspelled_name_df, threshold = 3, n_bands = 100, band_width = 1) %>%
         arrange(id_1, id_2)
 
@@ -266,4 +265,3 @@ test_that("argument `progress` works correctly", {
     "generating"
   )
 })
-

tests/testthat/test-test_lsh_properties.R

@@ -31,11 +31,11 @@ test_that("jaccard_hyper_grid_search validates inputs are length 1", {
 })
 
 test_that("hamming_probabilitiy gives accurate results", {
-    expect_equal(hamming_probability(0,10,10,1),1)
+  expect_equal(hamming_probability(0, 10, 10, 1), 1)
 
-    expect_equal(hamming_probability(1,10,1,1),.9)
-    expect_equal(hamming_probability(1,10,2,1),.99)
-    expect_equal(hamming_probability(1,10,3,1),.999)
+  expect_equal(hamming_probability(1, 10, 1, 1), .9)
+  expect_equal(hamming_probability(1, 10, 2, 1), .99)
+  expect_equal(hamming_probability(1, 10, 3, 1), .999)
 
-    expect_equal(hamming_probability(10,10,10,1),0)
+  expect_equal(hamming_probability(10, 10, 10, 1), 0)
 })