gistfile1.rmd

---
title: "Data/Structure Validation"
author: "agstudy"

---

This post is an answer to [SO question](http://stackoverflow.com/questions/30844363/data-structure-validation-for-r#comment49735091_30844363) about creating a well typed data structure in R.

I think, that in R the only way to define a typed data structure is to `S4 class`.  I should not that even S4 classes are *not strongly typed* since you can define your slot as `list`.

I Create an S4 class **TypedData**:

* I define 4 slots . I set the type of each slot. S4 class will validate gracefully for us that created object respect this typing
* Then I add a validation part to check our slots against  some conditions. Here for example age and weight should be positives values.
* You can also add some slot's default values.

```{r}
# Create TypedData class
TypedData <- setClass(
  # Set the name for the class
  "TypedData",
  
  # Define the slots
  representation (
    date    = "Date",
    age     = "numeric",
    weight  = "numeric",
    job     = "character"
    
  ),
  # Set the default values for the slots
  prototype=list(
    job = NA_character_
  ),
 
  # Make a function that can test to see if the data is consistent.
  validity=function(object)
  {
    if(length(object@age) ==0 || any(object@age < 0) )
      return("Age should be >0")
    if(length(object@weight) ==0 || any(object@weight < 0) )
      return("Weight should be >0")
    

    return(TRUE)
  }
)


```

Implement the S3 method to convert the S4 class to a `data.frame`

```{r}

as.data.frame.TypedData  <-
  function(x, row.names=NULL, optional=FALSE, ...)
  {
    value <- setNames(
      lapply(slotNames(x),function(y){
        col <- slot(x,y)
        if(length(col)<length(x@date))
          col <- rep(NA_character_,length(x@date))
        col
      }),
      slotNames(x))
    attr(value, "row.names") <- 
      as.character(seq_len(length(x@date)))
    class(value) <- "data.frame"
    value
  }


```

Create some data.
I Use the vector interpretation of the S4 class to create a data example.
It is is better for performance to manipulate columns.

```{r}
pers <- new("TypedData",  
            date=seq(as.Date("2015/1/1"), as.Date("2015/3/1"), "months"),
            age=c(20,30,50),
            weight=c(80,50,64))

pers
```

Now we convert the S4 object to a data.frame. 
We  check that the result is well typed.
```{r}
as.data.frame(pers)
str(as.data.frame(pers))
```