如何使这个嵌套的 for 循环更快地工作

Question

我的数据如下图：

txt$txt:

my friend stays in adarsh nagar
I changed one apple one samsung S3 n one sony experia z.
Hi girls..Friends meet at bangalore
what do u think of ccd at bkc

我有一份详尽的城市名称列表。下面列出其中的几个：

city:

ahmedabad
adarsh nagar
airoli
bangalore
bangaladesh
banerghatta Road
bkc
calcutta

我正在txt$txt 中搜索城市名称（来自我拥有的“城市”列表），如果它们存在，则将它们提取到另一列中。所以下面的简单循环对我有用……但在更大的数据集上需要很多时间。

for(i in 1:nrow(txt)){
    a <- c()
    for(j in 1:nrow(city)){
        a[j] <- grepl(paste("\\b",city[j,1],"\\b", sep = ""),txt$txt[i])        
    }
    txt$city[i] <- ifelse(sum(a) > 0, paste(city[which(a),1], collapse = "_"), "NONE")  
}   

我尝试使用 apply 函数，这是我能达到的最大值。

apply(as.matrix(txt$txt), 1, function(x){ifelse(sum(unlist(strsplit(x, " ")) %in% city[,1]) > 0, paste(unlist(strsplit(x, " "))[which(unlist(strsplit(x, " ")) %in% city[,1])], collapse = "_"), "NONE")})
[1] "NONE"      "NONE"      "bangalore" "bkc"  

Desired Output:
> txt
                                                       txt         city
1                          my friend stays in adarsh nagar adarsh nagar
2 I changed one apple one samsung S3 n one sony experia z.         NONE
3                      Hi girls..Friends meet at bangalore    bangalore
4                            what do u think of ccd at bkc          bkc    

我希望在 R 中实现更快的进程，它的作用与上面的 for 循环相同。请指教。谢谢

Answer 1

试试这个：

# YOUR DATA
##########
txt <- readLines(n = 4)
my friend stays in adarsh nagar and airoli
I changed one apple one samsung S3 n one sony experia z.
Hi girls..Friends meet at bangalore
what do u think of ccd at bkc

city <- readLines(n = 8)
ahmedabad
adarsh nagar
airoli
bangalore
bangaladesh
banerghatta Road
bkc
calcutta

# MATCHING
##########
matches <- unlist(setNames(lapply(city, grep, x = txt, fixed = TRUE), 
                           city))
(res <- (sapply(1:length(txt), function(x) 
  paste0(names(matches)[matches == x], collapse = "___"))))
# [1] "adarsh nagar___airoli" ""                      
# [3] "bangalore"             "bkc" 

Answer 2

这应该更快：

bigPattern <- paste('(\\b',city[,1],'\\b)',collapse='|',sep='')
txt$city <- sapply(regmatches(txt$txt,gregexpr(bigPattern,txt$txt)),FUN=function(x) ifelse(length(x) == 0,'NONE',paste(unique(x),collapse='_')))

说明：

在第一行中，我们构建了一个匹配所有城市的大正则表达式，例如：

(\\bahmedabad\\b)|(\\badarsh nagar\\b)|(\\bairoli\\b)| ...

然后我们将gregexpr与regmatches结合使用，这样我们就得到了txt$txt中每个元素的匹配列表。

最后，使用简单的sapply，对于列表中的每个元素，我们连接匹配的城市（在删除重复项后，即多次提及的城市）。

Answer 3

这是使用stringi 包中的stri_extract_first_regex 的可能性：

library(stringi)

# prepare some data
df <- data.frame(txt = c("in adarsh nagar", "sony experia z", "at bangalore"))
city <- c("ahmedabad", "adarsh nagar", "airoli", "bangalore")

df$city <- stri_extract_first_regex(str = df$txt, regex = paste(city, collapse = "|"))

df
#               txt         city
# 1 in adarsh nagar adarsh nagar
# 2  sony experia z         <NA>
# 3    at bangalore    bangalore