write_rule: Write new rule to be verified.
In kubistmi/verifier: Data verification (validation)

Description Usage Arguments Details Value Author(s) Examples

write_rule is a function to help you write new rules. Can write several rules at once using vectors.

1	write_rule(name, x, type, y, def , result, na.rm = "", multiple)

`name`	character. Name of the rule. Needs to be unique for every rule.
`x`	object. Data against which the rule is to be verified.
`type`	character. Possible types are: `integ` checks for presence of all unique values of x in y `summary` checks summarised data against original, more detailed data `na` checks the missing values `def` any other type, needs to be defined in def parametr.
`y`	vector. Used in `type` `integrity` (y is the reference table), `summary` (y is the summarised table), or in some cases of `def`.
`def`	character. Used for user defined rules using any function in environment, accepts only the name of the function, e.g. "sum".
`result`	numeric. Used for `type` `numeric`, `def`, or `summary` to specify bounds, expected value or accuracy for `summary`, check Details for more explanation.
`na.rm`	logical or empty (""). Should the `NA` values be ommitted in computations of summaries and user defined functions? Empty, if the function does not use na.rm argument (or it should use default value).
`multiple`	logical. Are multiple rules being written? If TRUE, then x (and y, if used) should be the names of the objects (not the objects themself)

You can create several rules at once assigning vectors into the parametrs and specifying multiple = TRUE. When the length of any parameter is longer than one the function uses cbind to create a data.frame (will repeat values of vectors with smaller size). Names of the rules need to be unique.
Special requirements:

summary Both x and y are expected to be a data.frame/table object either with columns 'id' and 'value', or with two columns only (in this case, first column is expected to be id and second to contain value).
def The defined function must have a name (you cannot use operators such as <). If you want to use two parameters (not only x) in the function (such as rewriting operator as named function function(a,b){a < b}), then you need to specify the other parameter as y. If you want to compare a result of a function with some predefined value (suck as mean(x) == number), you need to specify the parameter result as this number. If the function output is vector with lenght > 1, it is expected to contain logical values and will be summarized (not compared with parameter result) for the final output..

There is a special use of parameter result in case of type = summary, in this case result can be used to define accuracy (tolerance of difference between values). Default tolerance is 0 (compared using ==.)

Returns a message confirming the creation of new rule.

Michal Kubista

## Not run: 
write_rule(name = "test1", x = mtcars$mpg, type = "def", def = "mean",
           result = 19.2, na.rm = TRUE)
# verify, that the mean  of mtcars$mpg is 19.2, omitting the missing values

write_rule(name = "integrity1", x = data$cities, type = "integ", y = "ref$cities")
# verify, that all of the cities used in the data are present in the reference table

## End(Not run)