SimilaR (version 1.0.6)

SimilaR_fromDirectory: Quantify Similarity of All Pairs of Functions

Description

An implementation of the SimilaR algorithm - a novel method to quantify the similarity of R functions based on program dependence graphs. Possible use cases include detection of code clones for improving software quality and of plagiarism among students' homework assignments.

SimilaR_fromDirectory scans for function definition in *.R files in a given directory and performs pairwise comparisons.

Usage

SimilaR_fromDirectory(dirname, returnType = c("data.frame", "matrix"),
  fileTypes = c("function", "file"), aggregation = c("tnorm", "sym",
  "both"))

Arguments

dirname

path to a directory with *.R source files

returnType

"data.frame" or "matrix"; indicates the output object type

fileTypes

"function" or "file"; indicates which pairs of functions extracted from the source files in dirname should be compared; "function" compares each function against every other function; "file" compares only the functions defined in different source files

aggregation

"sym", "tnorm", or "both"; specifies which model of similarity asymmetry should be used; "sym" means that one (overall) value of similarity is computed; "both" evaluates and returns the degree to which the first function in a function pair is similar ("contained in", "is subset of") to the second one, and, separately, the extent to which the second function is similar to the first one; "tnorm" computes two similarity values and aggregates them to a single number

Value

If returnType is equal to "data.frame", a data frame that gives the information about the similarity of the inspected pairs of functions, row by row, is returned. Columns of the data frame are as follows:

  • name1 - the name of the first function in a pair

  • name2 - the name of the second function in a pair

  • SimilaR - values in the [0,1] interval as returned by the SimilaR algorithm; 1 denotes that the functions are equivalent, while 0 means that they are totally dissimilar; if aggregation is equal to "both", two similarity values are given: the one with suffix "12", which means how much the first function is a subset of the second, and the another one with suffix "21" which means how much the second function is a subset of the first one

  • decision - 0 or 1; 1 means that two functions are classified as similar, and 0 otherwise.

Rows in the data frame are sorted with respect to the SimilaR column (descending).

If returnType is equal to "matrix", a square matrix is returned. The element at index (i,j) equals to the similarity degree between the i-th and the j-th function. When aggregation is equal to "sym" or "tnorm", the matrix is symmetric. Column names and row names of the matrix are names of the compared functions.

Details

Note that, depending on the "aggregation" argument, the method may either return a single value, representing the overall (symmetric) similarity between a pair of functions, or or two different values, measuring the (non-symmetric) degrees of "subsethood". The user might possibly wish to aggregate these two values by means of some custom aggregation function.

References

Bartoszuk M., A source code similarity assessment system for functional programming languages based on machine learning and data aggregation methods, Ph.D. thesis, Warsaw University of Technology, Warsaw, Poland, 2018.

Bartoszuk M., Gagolewski M., Binary aggregation functions in software plagiarism detection, In: Proc. FUZZ-IEEE'17, IEEE, 2017.

Bartoszuk M., Beliakov G., Gagolewski M., James S., Fitting aggregation functions to data: Part II - Idempotentization, In: Carvalho J.P. et al. (Eds.), Information Processing and Management of Uncertainty in Knowledge-Based Systems, Part II (Communications in Computer and Information Science 611), Springer, 2016, pp. 780-789. doi:10.1007/978-3-319-40581-0_63.

Bartoszuk M., Beliakov G., Gagolewski M., James S., Fitting aggregation functions to data: Part I - Linearization and regularization, In: Carvalho J.P. et al. (Eds.), Information Processing and Management of Uncertainty in Knowledge-Based Systems, Part II (Communications in Computer and Information Science 611), Springer, 2016, pp. 767-779. doi:10.1007/978-3-319-40581-0_62.

Bartoszuk M., Gagolewski M., Detecting similarity of R functions via a fusion of multiple heuristic methods, In: Alonso J.M., Bustince H., Reformat M. (Eds.), Proc. IFSA/EUSFLAT 2015, Atlantis Press, 2015, pp. 419-426.

Bartoszuk M., Gagolewski M., A fuzzy R code similarity detection algorithm, In: Laurent A. et al. (Eds.), Information Processing and Management of Uncertainty in Knowledge-Based Systems, Part III (CCIS 444), Springer-Verlag, Heidelberg, 2014, pp. 21-30.

See Also

Other SimilaR: SimilaR_fromTwoFunctions

Examples

Run this code
# NOT RUN {
## Typical example, where we wish to compare the functions from different files,
## but we do not want to compare the functions from the same file.
## There will be one value describing the overall similarity level.
SimilaR_fromDirectory(system.file("testdata","data",package="SimilaR"),
                                 returnType = "data.frame",
                                 fileTypes="file",
                                 aggregation = "sym")

## In this example we want to compare every pair of functions: even those
## defined in the same file. Two (non-symmetric) similarity degrees
## are output.
SimilaR_fromDirectory(system.file("testdata","data2",package="SimilaR"),
                      returnType = "data.frame",
                      fileTypes="function",
                      aggregation = "both")


# }

Run the code above in your browser using DataCamp Workspace