[R] Date read correctly from CSV, then reformatted incorrectly by R

Richard O'Keefe r@oknz @end|ng |rom gm@||@com
Mon Nov 22 03:54:30 CET 2021

CSV data is very often strangely laid out.  For analysis,
Buffer Date Reading
100    ...  ...
100    ...  ...
and so on is more like what a data frame should be.  I get
quite annoyed when I finally manage to extract data from a
government agency only to find that my tax money has been
spent on making it harder to access than it needed to be.

(1) You do NOT need any additional library to convert dates.
    ?strptime is quite capable.

(2) Just because reshaping CAN be done in R doesn't mean it
    SHOULD be.  Instead of reading data in as the wrong format
    and then hacking it into shape every time, it makes sense
    to convert the data once and only once, then load the
    converted data.  It took just a couple of minutes to write

      (CSVDecoder read: 'transpose-in.csv') bindOwn: [:source |
        (CSVEncoder write: 'transpose-out.csv') bindOwn: [:target |
          source next bind: [:header | "Label date-1 ... date-n"
            target nextPut: {header first. 'Date'. 'Reading'}.
            [source atEnd] whileFalse: [
              source next bind: [:group |
                group with: header keysAndValuesDo: [:index :reading :date |
                  1 < index ifTrue: [
                    (date subStrings: '/') bind: [:dmy |
                      (dmy third,'-',dmy second,'-',dmy first) bind: [:iso |
                        target nextPut: {group first. iso.

    in another programming language, run it, and turn your example into
   You could do the same kind of thing easily in Perl, Python, F#, ...
   Then just read the table in using
   read.csv("transpose-out.csv", colClasses = c("integer","Date","numeric"))
   and you're away laughing.

(3) Of course you can do the whole thing in base R.

    h <- read.csv("transpose-in.csv", header=FALSE, nrows=1,
    d <- strptime(h[1,-1], format="%d/%m/%Y")
    b <- read.csv("transpose-in.csv", header=FALSE, skip=1)
    r <- expand.grid(Date=d, Buffer=b[,1])
    r$Result <- as.vector(t(as.matrix(b[,-1])))

    (A) You don't have to read a CSV file (or any other) all in one piece.
        This pays off when the structure is irregular.
    (B) You don't HAVE to accept or convert column names.
    (C) strptime is your friend.
    (D) expand.grid is particularly handy for "matrix form" CSV data.
    (E) Someone who suggests doing something in another language because
        it is easier can end up with egg on his face when doing the whole
        thing in R turns out to be easier, simpler, and far more obvious.
        (A) really is an important lesson.
    (F) It's *amazing* what you can do in base R.  It is useful to
        familiarise yourself with its capabilities before considering other
        packages.  Compositional data?  Not in base R.  Correspondence
        analysis?  Not in base R.  Data reshaping?  Very much there.

On Sun, 21 Nov 2021 at 06:09, Philip Monk <prmonk using gmail.com> wrote:

> Hello,
> Simple but infuriating problem.
> Reading in CSV of data using :
> ```
> # CSV file has column headers with date of scene capture in format
> dd/mm/yyyy
> # check.names = FALSE averts R incorrectly processing dates due to '/'
> data <- read.csv("C:/R_data/Bungala (b2000) julian.csv", check.names =
> # Converts data table from wide (many columns) to long (many rows) and
> creates the new object 'data_long'
> # Column 1 is the 'Buffer' number (100-2000), Columns 2-25 contain monthly
> data covering 2 years (the header row being the date, and rows 2-21 being a
> value for each buffer).
> # Column headers for columns 2:25 are mutated into a column called 'Date',
> values for each buffer and each date into the column 'LST'
> data_long <- data %>% pivot_longer(cols = 2:25, names_to = "Date",
> values_to = "LST")
> # Instructs R to treat the 'Date' column data as a date
> data_long$Date <- as.Date(data_long$Date)
> ```
> Using str(data), I can see that R has correctly read the dates in the
> format %d/%m/%y (e.g. 15/12/2015) though has the data type as chr.
> Once changing the type to 'Date', however, the date is reconfigured.  For
> instance, 15/01/2010 (15 January 2010), becomes 0015-01-20.
> I've tried ```data_long$Date <- as.Date(data_long$Date, format =
> "%d/%m.%y")```, and also ```tryformat c("%d/%m%y")```, but either the error
> persists or I get ```NA```.
> How do I make R change Date from 'chr' to 'date' without it going wrong?
> Suggestions/hints/solutions would be most welcome.  :)
> Thanks for your time,
> Philip
> Part-time PhD Student (Environmental Science)
> Lancaster University, UK.
> ~~~~~
> I asked a question a few weeks ago and put together the answer I needed
> from the responses but didn't know how to say thanks on this list.  So,
> thanks Andrew Simmons, Bert Gunter, Jeff Newmiller and Daniel Nordlund!
>         [[alternative HTML version deleted]]
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

	[[alternative HTML version deleted]]

More information about the R-help mailing list