Additional feature - Conversions dialog or elsewhere

[shadrackkibet]

I have stations catalogue and I would like to map all stations at KMD. Below is a section of the data.

I would like to have a facility in R-Instat to convert Degrees and minutes + seconds(when available) to latitude and longitudes.

This is possible in R using dms2dd() function from biogeo package. Using str_extract() from stringr package I extract letters (N, S, E, W) and minutes of latitude or longitude from Minutes columns. This is already possible using Find Replace dialog . You need to know some regex to be able to give the correct pattern. How can this be made easy for users who are not familiar with Regex? Here is a section of result columns.

Another alternative is using measurements::conv_unit() but this doesn't deal with "E/W" "N/S" entries, it requires positive/negative degrees.

I suggest this as another option in the conversions dialog or we can have this as a special dialog under mapping section? @rdstern @dannyparsons what are your views?

[rdstern]

It would be good to have this. If I have understood you correctly then we can take out the N, S, E, W using Prepare > Column: Text > Transform > Substring. Then doing from 1 to -2 will get the number part and from -2 to -1 will get the direction. Then I have used the Calculate simply as degrees + minutes/60, but it would be better to have an extra option on the conversions dialogue, because that would also cope better with the N/S question. Nice example. I wonder how many countries follow the KMD example. It isn't "tidy" to mix the minutes and the N/S or E/W? I would also like eventually to include regex as well! Cheers Roger Roger Stern Statistical Services Centre University of Reading, Whiteknights Road, Reading RG6 6FN Tel: +44 (0) 118 378 6421 (or 8025) e-mail r.d.stern@reading.ac.uk<mailto:r.d.stern@reading.ac.uk> web: www.reading.ac.uk/ssc<http://www.reading.ac.uk/ssc>

…

________________________________ From: Shadrack Kibet <notifications@github.com> Sent: 28 June 2019 05:25 To: africanmathsinitiative/R-Instat Cc: Subscribed Subject: [africanmathsinitiative/R-Instat] Additional feature - Conversions dialog or elsewhere (#5395) I have stations catalogue and I would like to map all stations at KMD. Below is a screenshot of the data. [image]<https://user-images.githubusercontent.com/26301973/60316175-32da4e80-9972-11e9-9f03-ee8dcee3a4f6.png> I would like to have a facility in R-Instat to convert Degrees and minutes + seconds(when available) to latitude and longitudes. This is possible in R using dms2dd() function from biogeo package. Using str_extract() from stringr package I extract letters (N, S, E, W) and minutes of latitude or longitude from Minutes columns. This is already possible using Find Replace dialog . You need to know some regex to be able to give the correct pattern. How can this be made easy for users who are not familiar with Regex? Here is a screenshot of result columns. [image]<https://user-images.githubusercontent.com/26301973/60316889-3b805400-9975-11e9-9fb9-74342b990432.png> Another alternative is using measurements::conv_unit() but this doesn't deal with "E/W" "N/S" entries, it requires positive/negative degrees. I suggest this as another option in the conversions dialog or we can have this as a special dialog under mapping section? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#5395>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACVU2ZNOFHMWMJG3H37TTPDP4WHETANCNFSM4H4CABXA>.

[dannyparsons]

We've discussed this with Shadrack and think the best approach is to write our own function so we can make a few additions to the dms2dd() function and also avoid having to add a whole package for a single function.

[shadrackkibet]

dms2dd <- function (dd, mm, ss, ns) {
  n <- match(toupper(ns), LETTERS)
  we1 <- n == 19 | n == 23
  sgn <- ifelse(we1 == T, -1, 1)
  mm1 <- ifelse(missing(mm), 0, mm)
  ss1 <- ifelse(missing(ss), 0, ss)
  if(missing(dd))  stop("degrees must be supplied")
  decdeg <- (dd + ((mm1 * 60) + (ss1))/3600) * sgn
  return(decdeg)
}

@dannyparsons here is the implementation of the R function. Any comments?

[shadrackkibet]

Here is the design.

[dannyparsons]

Looks pretty good. Few small points:

• We can add a check that ns only contains the right letters, currently I could have "y" and it wouldn't complain.
• we1 == T can be replaced with we1, since it's already logical
• instead of checking if mm and ss are missing I would have 0 as their defaults, then a check is not needed. That's then clearer to the user what is happening
• it would be good to have a slightly different name for the function than dms2dd to avoid a clash
• I think you want mm1 / 60 instead of *
• We can have a check that mm and ss are between 0 and 60
• If ns is provided then dd must be positive (I think that's right?) as another check.

I think adding these checks are important because its good quality control of the type of data that will be used for this.

Is it also common to not have the N/S E/W column and instead the values are already positive/negative? If so, ns can be optional.

[shadrackkibet]

Thanks, yes it is true that when ns is provided then dd should be positive. On your 5th point, I want / and not * because I have minutes converted to seconds.

convert_to_dec_deg <- function (dd, mm = 0 , ss = 0, ns) {
  n <- match(toupper(ns), LETTERS)
  if(!all(toupper(ns) %in% c("E","W","N","S"))) stop("ns must be one of E, W, N or S")
  if(missing(dd))  stop("dd must be supplied")
  if(!all(mm %in% 0:60)|!all(ss %in% 0:60)) stop("mm and ss must be between 0 and 60")
  if(!missing(ns)&any(dd<0)) stop("dd must be positive") 
  we1 <- n == 19 | n == 23
  sgn <- ifelse(we1, -1, 1)
  decdeg <- (dd + ((mm * 60) + (ss))/3600) * sgn
  return(decdeg)
}

I am not sure if it is common to have N/S E/W column. I think if this is not available then dd has positives\negatives. What do you think, should we also add a check if dd has positives and negatives then ns isn't needed?

I think we should also check that dd, mm, ss and ns are of the same length. I suggest we give a warning instead of stopping in this case.

[dannyparsons]

You don't want %in% 0:60 because that only allows integers.
I've simplified a bit. I don't think we get too complicated with the checks, if dd has negative values and dir is also supplied they should go back and decide what they want to do. I've modified it so that dir is not required. We don't have to use this if we don't want to.
A length check would be good but if any are length 1 then that shouldn't get any message, otherwise it should give an error since we can't calculate with different lengths.
We should also check it works with missing values - if any component is missing the result is missing.

convert_to_dec_deg <- function (dd, mm = 0 , ss = 0, dir) {
  if(missing(dd))  stop("dd must be supplied")
  if(!missing(dir)) {
    dir <- toupper(dir)
    if(!all(na.omit(dir) %in% c("E", "W", "N", "S"))) stop("dir must only contain direction letters E, W, N or S")
  if(any(dd < 0)) stop("dd must be positive if dir is supplied") 
  }
  if(!all(mm >= 0 & mm <= 60, na.rm = TRUE)) stop("mm must be between 0 and 60")
  if(!all(ss >= 0 & ss <= 60, na.rm = TRUE)) stop("ss must be between 0 and 60")
  sgn <- ifelse(is.na(dir), NA, ifelse(dir %in% c("S", "W"), -1, 1))
  decdeg <- (dd + ((mm * 60) + ss)/3600) * sgn
  return(decdeg)
}

[shadrackkibet]

Thanks.