Offensive Programming in action (part III)

[This article was first published on NEONIRA, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

This is the third post on offensive programming, dedicated to using offensive programming (OP).

You may refer to easy introduction to offensive programming to acquire some basic knowledge about offensive programming, and to discover offensive programming to get access to online materials.

Let’s see how OP helps in defining function with semantic naming for parameter names, and in providing expected function return type. This will make use of many tools provided by package wyz.code.offensiveProgramming.

Focus on a given type suffix

Let’s consider you want to know more about type suffix ’s’. Here is a typical session you could use.

library('wyz.code.offensiveProgramming')

# get the default type factory
tf <- retrieveFactory()

# is 's' a recorded type suffix?
tf$checkSuffix('s')

## [1] TRUE

# what means 's' type suffix
tf$getType('s')

## [1] "string"

# get verification function for type suffix 's'
tf$getVerificationFunction('s')

## function (o_1_) 
## {
##     if (!is.character(o_1_)) 
##         return(FALSE)
##     if (length(o_1_) == 0) 
##         return(TRUE)
##     all(is.na(o_1_) == FALSE)
## }
## <bytecode: 0x55a7a37500a0>
## <environment: 0x55a7a3732340>

# get type factory information for 'pi'
tf$getRecordedTypes()[suffix == 's']

##    suffix   type verify_function category
## 1:      s string      <function>    basic

A real use case

Raw approach

Let’s consider following function that helps organizing seats for your friend around the table (under no constraint indeed).

organizeTable <- function(guestNames_s) {
  guestNames_s[order(runif(length(guestNames_s)))]
}

guests <- c('Marie', 'Yves-Marie', 'Fabien', 'Marcello', 'Tina')
organizeTable(guests)

## [1] "Marcello"   "Marie"      "Tina"       "Yves-Marie" "Fabien"

organizeTable(character(0))

## character(0)

From a R perspective, obvuously yes. From OP perspective, answer is more subtle and depends on evaluation modes you choose.

Evaluation modes

Currently, 3 evaluation modes are supported. They are incremental modes.

  1. standard_R_evaluation mode,
  2. enhanced_R_evaluation mode, adds function return type evaluation to previous mode,
  3. type_checking_enforcement mode, adds function parameter type verification to previous mode.
# Available evaluation modes
defineEvaluationModes()

## [1] "standard_R_evaluation"     "enhanced_R_evaluation"    
## [3] "type_checking_enforcement"

Standard R evaluation mode

em <- EvaluationMode('standard_R_evaluation')
efrt <- 'x_s' # expected function returned type
runTransientFunction(organizeTable, list(guests), em, efrt)

## $status
## [1] TRUE
## 
## $value
## [1] "Tina"       "Yves-Marie" "Marcello"   "Fabien"     "Marie"     
## 
## $mode
## [1] "standard_R_evaluation"

Using, standard_R_evaluation mode brings status, value and mode results. It does not provide any extraneous information than a call in your R console. This mode is provided to ease comparisons. Not to be used solely, indeed.

Enhanced evaluation mode

This mode checks function returned value against expected function return type.

em <- EvaluationMode('enhanced_R_evaluation')
runTransientFunction(organizeTable, list(guests), em, efrt)

## $status
## [1] TRUE
## 
## $value
## [1] "Tina"       "Marcello"   "Marie"      "Fabien"     "Yves-Marie"
## 
## $mode
## [1] "enhanced_R_evaluation"
## 
## $function_return_type_check
##    parameter_name                       parameter_value validity
## 1:            x_s Tina,Marcello,Marie,Fabien,Yves-Marie     TRUE
##                message
## 1: good type in values

Using, enhanced_R_evaluation mode brings function_return_type results, as a data.table. The validity column provides information about type concordance between resulting value and expected type.

Enhanced evaluation mode

This mode checks parameter values against parameter semantic name specifications.

em <- EvaluationMode('type_checking_enforcement')
runTransientFunction(organizeTable, list(guests), em, efrt)

## $status
## [1] TRUE
## 
## $value
## [1] "Yves-Marie" "Tina"       "Marcello"   "Fabien"     "Marie"     
## 
## $mode
## [1] "type_checking_enforcement"
## 
## $parameter_type_checks
##    parameter_name                       parameter_value validity
## 1:   guestNames_s Marie,Yves-Marie,Fabien,Marcello,Tina     TRUE
##                message
## 1: good type in values
## 
## $function_return_type_check
##    parameter_name                       parameter_value validity
## 1:            x_s Yves-Marie,Tina,Marcello,Fabien,Marie     TRUE
##                message
## 1: good type in values

Using, type_checking_enforcement mode brings parameter_type_checks results, as a data.table. Here also, the validity column provides information about type concordance between provided parameter value and parameter semantic name specification. Detail level is one line for each function parameter, to ease discovery of and remediance to uncompliances.

Let’s check them

em <- EvaluationMode('type_checking_enforcement')
runTransientFunction(organizeTable, list(guests), em, efrt)

## $status
## [1] TRUE
## 
## $value
## [1] "Tina"       "Fabien"     "Yves-Marie" "Marcello"   "Marie"     
## 
## $mode
## [1] "type_checking_enforcement"
## 
## $parameter_type_checks
##    parameter_name                       parameter_value validity
## 1:   guestNames_s Marie,Yves-Marie,Fabien,Marcello,Tina     TRUE
##                message
## 1: good type in values
## 
## $function_return_type_check
##    parameter_name                       parameter_value validity
## 1:            x_s Tina,Fabien,Yves-Marie,Marcello,Marie     TRUE
##                message
## 1: good type in values

runTransientFunction(organizeTable, list(character(0)), em, efrt)

## $status
## [1] TRUE
## 
## $value
## character(0)
## 
## $mode
## [1] "type_checking_enforcement"
## 
## $parameter_type_checks
##    parameter_name parameter_value validity             message
## 1:   guestNames_s                     TRUE good type in values
## 
## $function_return_type_check
##    parameter_name parameter_value validity             message
## 1:            x_s                     TRUE good type in values

From R point of view, both calls are legal. From offensive programming, only calls returning a status TRUE are valid.

Accepting or not the second case is indeed a matter of input scope specification. Does organizing a table for no person have any sense?

If your answer is yes, then keep this function definition. If no, how could you improve previous implementation?

Refined approach

We will change the parameter name and specify length constraint on it. Note that function body is semantically exactly the same as for previous function. No change in implementation algorithm. Just parameter renaming propagation changes are applied.

organizeTableBis <- function(guestNames_s_1m) {
  guestNames_s_1m[order(runif(length(guestNames_s_1m)))]
}

organizeTableBis(guests)

## [1] "Fabien"     "Tina"       "Marie"      "Yves-Marie" "Marcello"

organizeTableBis(character(0))

## character(0)

runTransientFunction(organizeTableBis, list(guests), em, efrt)

## $status
## [1] TRUE
## 
## $value
## [1] "Tina"       "Fabien"     "Marcello"   "Yves-Marie" "Marie"     
## 
## $mode
## [1] "type_checking_enforcement"
## 
## $parameter_type_checks
##     parameter_name                       parameter_value validity
## 1: guestNames_s_1m Marie,Yves-Marie,Fabien,Marcello,Tina     TRUE
##                message
## 1: good type in values
## 
## $function_return_type_check
##    parameter_name                       parameter_value validity
## 1:            x_s Tina,Fabien,Marcello,Yves-Marie,Marie     TRUE
##                message
## 1: good type in values

runTransientFunction(organizeTableBis, list(character(0)), em, efrt)

## $status
## [1] FALSE
## 
## $value
## character(0)
## 
## $mode
## [1] "type_checking_enforcement"
## 
## $parameter_type_checks
##     parameter_name parameter_value validity
## 1: guestNames_s_1m                    FALSE
##                                       message
## 1: wrong length, was expecting [1m] , got [0]
## 
## $function_return_type_check
##    parameter_name parameter_value validity             message
## 1:            x_s                     TRUE good type in values

Now the second case is flagged as invalid from OP point of view. One of the reasons is given explicitly by the parameter_type_check data.table.

To conclude …

You have

  1. been introduced to semantic naming for function parameter names,
  2. been sensitized to semantic naming length specification impact,
  3. been shown how and when to use predefined evaluation modes,
  4. been initiated to use common OP tools, runTransientFunction, EvaluationMode, and function parameter type factory operations,
  5. been trained to interpret evaluation results

Great, we are more than half the way. Next post will be about registering your own types, and managing your classes and their related functional tests. Stay tuned.

To leave a comment for the author, please follow the link and comment on their blog: NEONIRA.

R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)