Reading large data with messy strings and multiple string indicators R

Question

I have a large (8GB+) csv file (comma-separated) that I want to read into R. The file contains three columns

date #in 2017-12-27 format
text #a string
type #a label per string (either NA, typeA, or typeB)

The problem I encounter is that the text column contains various string indicators: ' (single quot. marks), " (double quot. marks), no quot. marks, as well as multiple separated strings.

E.g.

date        text                        type
2016-01-01  great job!                  NA
2016-01-02  please, type "submit"       typeA
2016-01-02  "can't see the "error" now" typeA
2016-01-03  "add \\"/filename.txt\\""   NA

To read these large data, I tried:

Base read.csv and readr's read_csv function: work fine for a portion but fail (probably due to memory) or take ages to read
chunking the data via Mac terminal into batches of 1m lines: fails because lines seem to break arbitrarily
Using fread (preferred as I hope this will solve the two other issues): fails with Error: Expecting 3 cols, but line 1103 contains text after processing all cols.

My idea is to work around these issues by using specifics of the data that I know, i.e. that each line starts with a date and ends with either NA, typeA, or typeB.

How could I implement this (either using pure readLines or into fread)?

Edit: Sample data (anonymized) as opened with Mac TextWrangler:

"date","text","type"
"2016-03-30","Maybe use `tapply` from `base`, and check how that works.",NA
"2016-04-01","Fiex this now. Please check.","typeA"
"2016-04-01","Does it work? Maybe try the other approach.","typeB"
"2016-04-01","This won't work. You should remove ABC ... each line starts with a date and ends with ... and this line is veeeeeeeeeeeeeeeeeery long.",NA
"2014-05-02","Tried to remove ""../"" but no success @myid",typeA

Sample data 2:

"date","text","type"
"2018-05-02","i try this, but it doesnt work",NA
"2018-05-02","Thank you very much. Cheers !!",NA
"2018-05-02","@myid. I'll change this.",NA

Sample data for reproducible fread error "Expecting 3 cols, but line 3 contains text after processing all cols.":

"date","text","type"
"2015-03-02","Some text, some text, some question? Please, some question?",NA
"2015-03-02","Here you have the error ""Can’t access {file \""Macintosh HD:abc:def:filename\"", \""/abc.txt\""} from directory."" something -1100 from {file ""Macintosh HD:abc:def:filename"", ""/abc.txt""} to file",NA
"2015-03-02","good idea",NA
"2015-03-02","Worked perfectly :)",NA

SessionInfo:

R version 3.5.0 (2018-04-23)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.5

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRlapack.dylib

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] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] data.table_1.10.4-3 readr_1.1.1        

loaded via a namespace (and not attached):
[1] compiler_3.5.0   assertthat_0.2.0 R6_2.2.2         cli_1.0.0       
[5] hms_0.4.2        tools_3.5.0      pillar_1.2.2     rstudioapi_0.7  
[9] tibble_1.4.2     yaml_2.1.19      crayon_1.3.4     Rcpp_0.12.16    
[13] utf8_1.1.3       pkgconfig_2.0.1  rlang_0.2.0

I think your problem is that you have commas in your `text` column. e.g. `please, type "submit"` and this is why `fread` fails. We will need to see the actual data in its row format. E.g., try opening it with notepad — David Arenburg, Jun 14 '18 at 07:50
@DavidArenburg Added data. Do you have ideas how to fix this? — ben_aaron, Jun 14 '18 at 09:56
I was able to read both of your examples with `fread` without any issues. Maybe try updating your data.table/R versions — David Arenburg, Jun 14 '18 at 10:29
Thanks. I've now found the part where it goes wrong and added data that reproduces the exact error with fread. What do you think is the issue? — ben_aaron, Jun 14 '18 at 10:51
I was able to read that one too without issues. Can you show your `sessionInfo()`? — David Arenburg, Jun 14 '18 at 10:59

Prem · Answer 1 · 2018-06-14T11:17:36.450

readLines approach could be

infile <- file("test.txt", "r")
txt <- readLines(infile, n = 1)
df <- NULL

#change this value as per your requirement
chunksize <- 1

while(length(txt)){
  txt <- readLines(infile, warn=F, n = chunksize)
  df  <- rbind(df, data.frame(date = gsub("\\s.*", "", txt),
                              text = trimws(gsub("\\S+(.*)\\s+\\S+$", "\\1", txt)),
                              type = gsub(".*\\s", "", txt),
                              stringsAsFactors = F))
  }

which gives

> df
        date                          text  type
1 2016-01-01                    great job!    NA
2 2016-01-02         please, type "submit" typeA
3 2016-01-02   "can't see the "error" now" typeA
4 2016-01-03 "add \\\\"/filename.txt\\\\""    NA

Sample data: test.txt contains

date        text                        type
2016-01-01  great job!                  NA
2016-01-02  please, type "submit"       typeA
2016-01-02  "can't see the "error" now" typeA
2016-01-03  "add \\"/filename.txt\\""   NA

Update: You can modify above code with below regex parser to parse another set of sample data

df  <- rbind(df, data.frame(date = gsub("\"(\\S{10}).*", "\\1", txt),
                            text = gsub(".*\"\\,\"(.*)\"\\,(\"|NA).*", "\\1", txt),
                            type = gsub(".*\\,|\"", "", txt),
                            stringsAsFactors = F))

Another set of sample data:

"date","text","type"
"2016-03-30","Maybe use `tapply` from `base`, and check how that works.",NA
"2016-04-01","Fiex this now. Please check.","typeA"
"2016-04-01","Does it work? Maybe try the other approach.","typeB"
"2016-04-01","This won't work. You should remove ABC ... each line starts with a date and ends with ... and this line is veeeeeeeeeeeeeeeeeery long.",NA
"2014-05-02","Tried to remove ""../"" but no success @myid","typeA"

Thanks @Prem. I've posted sample data. If I use my sample data 2 with your approach, I get lines like: `"2018-05-02","Thank[SEPARATEDHERE]you very much. Cheers[SEPARATEDHERE]!!",NA` — ben_aaron, Jun 14 '18 at 10:07

Reading large data with messy strings and multiple string indicators R

1 Answers1