Automatic Labeling of the Object-oriented Source Code: The Lotus Approach
0 likes28 views
The document presents the 'Lotus' approach for automatically labeling object-oriented software code to enhance understanding and documentation by extracting keywords from software identifiers. It discusses the methodology and results of applying this approach to case studies, demonstrating its effectiveness in retrieving accurate software labels. Future work aims to incorporate code comments and tag cloud visualization for improved documentation and analysis of software evolution.
![Sci.Int.(Lahore),30(1),45-48,2018 ISSN 1013-5316;CODEN: SINTE 8 45
January-February
AUTOMATIC LABELING OF THE OBJECT-ORIENTED SOURCE CODE:
THE LOTUS APPROACH
Ra'Fat Al-Msie'deen
Department of IT, Faculty of Information Technology, Mutah University, P.O. Box 7, Mutah 61710, Karak, Jordan
Email: rafatalmsiedeen@mutah.edu.jo
ABSTRACT: Most of open-source software systems become available on the internet today. Thus, we need automatic
methods to label software code. Software code can be labeled with a set of keywords. These keywords in this paper
referred as software labels. The goal of this paper is to provide a quick view of the software code vocabulary. This paper
proposes an automatic approach to document the object-oriented software by labeling its code. The approach exploits all
software identifiers to label software code. The paper presents the results of study conducted on the ArgoUML and
drawing shapes case studies. Results showed that all code labels were correctly identified.
Keywords: Software engineering, Software comprehension, Reverse engineering, Software visualization, Code Label.
1. INTRODUCTION
Program understanding is the main activity for software
maintenance and evolution [1]. The main idea of this paper
is to reduce the amount of information mined from the
software source code to provide a quick overview of the
software code vocabularies.
This paper proposes a novel approach called Lotus to
automatically provide labels for software code. Name of
the approach inspired by the Lotus flower. Lotus stands for
automatic Labeling of the ObjecT-oriented SoUrce code.
Several companies are facing problems with legacy
software systems such as: software understanding and
maintenance. The reason behind these problems is the
absence of software documentation [2]. The quick
development of software engineering approaches and the
increasing of software system complexity has led to a large
production of textual information contained in software
artifacts (e.g., software source code and design documents).
As a result, numerous papers investigated the analysis of
textual information contained in the software artifacts to
support software engineering activities [3] such as: feature
location [4] and software documentation [5].
To extract labels from the software source code, the Lotus
approach relies on the software identifiers (i.e., package,
class, attribute and method). Basically, the most important
words are included in the package, class, method, and
attribute names [6]. The extracted labels can give a hint to
the software developers about software code vocabularies.
The software code labeling process aims to analyze the
legacy software to determine its main keywords, where the
huge amounts of code need to be understood. To help a
human expert to document the legacy software, the Lotus
approach generates a set of labels represent the software
vocabularies.
This paper proposes an automatic approach which extracts
labels from software using its source code. Compared with
the existing work that labeling software source code (cf.
section 2), the novelty of the Lotus approach is that it
exploits all software identifiers to generate software labels
in an efficient way.
Lotus approach identifies the software code using static
code parser. Then, the approach identifies all software
identifiers. Then, Lotus splits the name of identifier into a
set of keywords based on the CamelCase methods [7].
Then, Lotus returns each keyword to its stem based on the
WordNet1
. Finally, Lotus generates software labels. A
1
WordNet: https://wordnet.princeton.edu/
sample execution of the Lotus approach is displayed on
Figure 1.
Figure 1. Running the Lotus approach on a simple example.
Lotus approach is detailed in the remainder of this paper as
follows. Section 2 discusses the related work. Section 3
shows an overview of Lotus approach. Section 4 presents
the source code labeling process step by step. Section 5
describes the experimentation, while section 6 concludes
and provides perspectives for this work.
2. RELATED WORK
This section presents the related work and provides a
concise summary of the different approaches.
Kuhn [8] presents a lexical approach that uses the log-
likelihood ratios of word frequencies to automatically
provide labels for software components.
AL-Msie'Deen [5] developed a tool called Vsound to
visualize the software code and its dependencies. Vsound
relies on software identifiers to visualize and document the](https://image.slidesharecdn.com/1803-230126152422-7d6a2942/85/Automatic-Labeling-of-the-Object-oriented-Source-Code-The-Lotus-Approach-1-320.jpg)
![46 ISSN 1013-5316;CODEN: SINTE 8 Sci.Int.(Lahore),30(1),45-48,2018
January-February
software code. While Lotus approach extracts labels from
software identifiers.
AL-Msie'Deen et al. [9] proposed an approach called
REVPLINE to retrieve labels for the mined features from
the source code of software family based on the use-case
diagrams. REVPLINE gives as output for each feature
implementation, a label based on the use-case label [10].
Lucia et al. [6] suggest a method for source code labeling,
based on information retrieval techniques, to identify
relevant words in the source code. They applied numerous
information retrieval methods [1] to extract terms from the
names of software classes. Their approach does not
consider the names of packages, attributes and methods.
Lotus approach considers all granularity levels of software
code.
Kuhn et al. [11] suggest a method to cluster software
classes that use similar vocabulary together. Then, they use
latent semantic indexing to automatically label clusters
with their most relevant terms. Lotus approach aims to
document the software code as a collection of labels.
Most existing approaches are designed to extract labels to
name software features, components or clusters. In the
literature, there is no work identifies code labels at all
granularity levels. Also, many approaches manually extract
software labels. Conversely, Lotus is designed to
automatically extract software labels based on all software
identifiers.
3. APPROACH OVERVIEW
This section provides an overview of code labeling process
and describes the example that illustrates the remaining of
the paper.
The main goal of Lotus approach is to understand the
software code vocabularies. The Lotus approach aims to
provide the software developers with software labels. The
complex software system involves a huge amount of source
code (i.e., lines of code), so there are needs to deal with all
granularity levels of source code in an abstract way.
Figure 2. The source code labeling process.
The source code labeling process takes the software's
source code as its inputs and creates code labels as its
outputs. The goal of these labels is to provide us with a
quick overview of software vocabularies. The software
identifier names play a significant role in software
understanding, particularly when the software documents
are missing, or when the software code is very complex
such that the identifier names would tell more about code to
the software developers [6].
Figure 2 shows the source code labeling process in general.
Lotus approach extracts software code. Then, the approach
identifies software identifiers. Then, the approach splits the
software identifiers into keywords based on the CamelCase
method. Then, the Lotus returns each keyword to the word
root via WordNet. Finally, Louts builds software labels.
As an illustrative example, Lotus considers the drawing
shapes software [12]. The software allows user to draw
three types of shapes which are lines, rectangles and ovals2
.
This software used to better explain the code labeling
process.
4. SOURCE CODE LABELING PROCESS
This section describes the source code labeling process step
by step. Lotus approach identifies the source code labels in
five steps as detailed in the following.
4.1 Extracting source code
The Lotus approach only uses the software source code as
input of the labeling process and thus Lotus does not know
the code labels in advance. The first step of the labeling
process aims to extract the software source code. The Lotus
approach extracts all essential information from source
code such as: software identifiers and code dependencies
[5]. The static code parser3
takes software code to generate
the code file as output. The code file contains software
identifiers. The Lotus source code parser is written
completely in Java.
4.2 Identifying software identifiers
To comprehend the source code of legacy software
systems, it's important to work with all granularity levels of
the source code. Lotus approach provides us with four
kinds of documents (aka identifiers file). Each document
contains a set of identifier names. This step generates four
documents which are package, class, attribute and method
document.
Table 1. Examples of software identifiers from drawing shapes
software.
Package Names Class Names
Drawing MyLine
Drawing.Shapes MyOval
Drawing.Shapes.coreElements MyRectangle
Drawing.Shapes.coreFrame PaintJPanel
DrawingShapes
Examples of software identifiers are shown in Table 1. The
package document contains the names of all software
packages. The class document contains the names of all
software classes and so on.
4.3 Splitting the identifiers
The identifier names are used to define the main entities of
the software system (e.g., package and class). The names of
the identifiers represent a set of characters based on the
rules of programming languages [13].
Lotus relies on the CamelCase technique as an identifier
splitting algorithm. CamelCase method is a simple and
generally used method for identifier splitting algorithms [7]
2
Drawing shapes: https://sites.google.com/site/ralmsideen/tools
3
Static code parser: https://sites.google.com/site/ralmsideen/tools](https://image.slidesharecdn.com/1803-230126152422-7d6a2942/85/Automatic-Labeling-of-the-Object-oriented-Source-Code-The-Lotus-Approach-2-320.jpg)
![Sci.Int.(Lahore),30(1),45-48,2018 ISSN 1013-5316;CODEN: SINTE 8 47
January-February
and the rules of splitting are widely based on CamelCase
agreement. Lotus splits identifier names into a set of
keywords based on the camel-case syntax. For example, the
identifier DrawingShapes is split into two words: drawing
and Shapes.
Table 2. Examples of the camel-case identifier splitting
algorithm.
Identifier Name Token/word
Word-1 Word-2 Word-3
shapesArrayList shapes array list
setCurrentColor set current color
colorJButton color j button
createUserInterface create user interface
Examples of camel-case splitting algorithm are shown in
Table 2.
4.4 Stemming identifier keywords
Stemming is the procedure of stripping affixes (i.e.,
prefixes and suffixes) from words to form a stem [14].
Stemming is often applied to words in information retrieval
methods (e.g., latent semantic indexing [12]). The
stemming is used in Lotus to replace English words with
their root or stem [15].
Lotus returns each keyword to its root or word stem. For
instance, if we have the following keyword "drawing", the
root of this keyword will be "draw". After the stemming
was performed via WordNet the stems of keywords were
stored in the terms file to generate the software labels.
Table 3. A sample of the word stems from drawing shapes
software.
Identifier word Word stem
Pressed Press
Drawing Draw
Performed Perform
Dragged Drag
Table 3 shows a sample of the word stems (i.e., terms) from
drawing shapes software.
4.5 Building the software labels
In this step, Lotus approach builds software labels to show
a quick view of the used vocabularies in the software code.
Figure 3 shows the mined labels from drawing shapes
software.
Figure 3. The extracted labels from drawing shapes software.
In Lotus approach, the extracted labels called label map.
All information about source code vocabulary given in the
extracted label map. Thus, all code vocabularies can easily
be acquired when needed via Lotus approach.
5. EXPERIMENTATION
This section presents the experiment that conducted to
validate the Lotus approach. Firstly, this section presents
the ArgoUML case study. Then, it presents the source code
labeling outcomes and, at last, it presents the threats to
validity of the Lotus approach.
In addition to the toy example used in this paper (i.e.,
drawing shapes software), the Lotus approach has been
tested on the ArgoUML4
software system. ArgoUML is a
widely used open source tool for UML modeling tool [15]
and well known [16]. ArgoUML software represents a
large case study (i.e., 120,348 lines of code).
Figure 4. The extracted package label map from ArgoUML.
Figure 4 shows the extracted package label map from
ArgoUML software using Lotus prototype5
. The algorithm
execution time is equal to 50860 ms. The identified labels
(aka topics) are very helpful to document software
vocabularies. In Figure 4, the package labels presented in
an alphabetical order and without repetition.
Figure 5 shows the extracted class label map from
ArgoUML software. The software developer can easily
obtain all code labels using Lotus approach. For a lack of
studies that are evaluating the extracted labels from
software code, there was difficulty in evaluating the Lotus
approach.
Figure 5. The extracted class label map from ArgoUML.
The threat to the validity of Lotus is that camel-case syntax
may be not reliable in all cases to split software identifiers.
Also, Lotus considers only the Java software, and this
limits the Lotus implementation ability to deal only with
Java language. Also, if the label map comes up with labels
such as: m, l, get or in this does not tell a human expert
4
ArgoUML: http://argouml.tigris.org/
5
Lotus prototype: https://sites.google.com/site/ralmsideen/tools](https://image.slidesharecdn.com/1803-230126152422-7d6a2942/85/Automatic-Labeling-of-the-Object-oriented-Source-Code-The-Lotus-Approach-3-320.jpg)
![48 ISSN 1013-5316;CODEN: SINTE 8 Sci.Int.(Lahore),30(1),45-48,2018
January-February
much about the software system. In this case, the label map
was totally useless, since the developer knowledge is
missing [11].
Table 4: Summary of Lotus approach.
Objectives Programmed method
Code labeling √ Automatic √
Code understanding √ Input
Code documentation √ Software packages √
Code visualization √ Software classes √
Tool support Software attributes √
WordNet √ Software methods √
Evaluation Technique
Evaluation x Ad hoc algorithm √
Case study Camel-case algorithm √
Drawing shapes √ Splitting method
ArgoUML √ Camel-case √
Output
Package label map √
Class label map √
Attribute label map √
Method label map √
Table 4 summarizes Lotus approach where it presents the
objectives, programmed method, input, tool support,
evaluation, technique, case study, splitting method and
output.
6. CONCLUSION AND FUTURE WORK
This paper proposed an approach to extract code labels.
The goal of this paper is to document code vocabularies.
The novelty of this paper is that it exploits all software
identifiers to document software vocabularies. Lotus has
implemented on several case studies. Results showed that
all label maps were identified. Regarding future work,
Lotus plans to use code comments in the code labeling
process. Also, Lotus approach plans to exploit the tag cloud
visualization technique [17] to display label frequency
across software code. In addition, Lotus plans to study the
evolution of software labels in a set of software variants
using formal concept analysis [18].
7. REFERENCES
[1] A. D. Lucia, M. D. Penta, R. Oliveto, A.
Panichella, and S. Panichella, “Labeling source
code with information retrieval methods: an
empirical study,” Empirical Software Engineering,
19(5): 1383 – 1420, (2014).
[2] S. Souza, N. Anquetil, and K. Oliveira. “A study
of the documentation essential to software
maintenance,” SIGDOC '05, 68 – 75 (2005).
[3] R. Al-Msie’deen, A. Seriai, M. Huchard, C.
Urtado, S. Vauttier, and H. E. Salman, “Feature
location in a collection of software product
variants using formal concept analysis,” ICSR,
302 – 307, (2013).
[4] R. Al-Msie’deen, “Reverse engineering feature
models from software variants to build software
product lines: REVPLINE approach,” Ph.D.
dissertation, Montpellier 2 University, France,
(2014).
[5] R. Al-Msie’deen, “Visualizing object-oriented
software for understanding and documentation,”
IJCSIS, 13(5): 18 – 27, (2015).
[6] A. D. Lucia, M. D. Penta, R. Oliveto, A.
Panichella, and S. Panichella, “Using IR methods
for labeling source code artifacts: Is it
worthwhile?” ICPC, 193 – 202, (2012).
[7] B. Dit, L. Guerrouj, D. Poshyvanyk, and G.
Antoniol, “Can better identifier splitting
techniques help feature location?”, ICPC, 11 –
20, (2011).
[8] A. Kuhn, “Automatic labeling of software
components and their evolution using log-
likelihood ratio of word frequencies in source
code,” MSR, 175 – 178, (2009).
[9] R. Al-Msie’deen, A. Seriai, M. Huchard, C.
Urtado, and S. Vauttier, “Documenting the mined
feature implementations from the object-oriented
source code of a collection of software product
variants,” SEKE, 138 – 143, (2014).
[10] R. Al-Msie’deen, M. Huchard, A. Seriai, C.
Urtado, and S. Vauttier, “Automatic
documentation of [mined] feature
implementations from source code elements and
use-case diagrams with the REVPLINE
approach,” IJSEKE, 24(10): 1413 – 1438, (2014).
[11] A. Kuhn, S. Ducasse, and T. Gîrba, “Semantic
clustering: Identifying topics in source code,”
Information & Software Technology, 49(3): 230 –
243, (2007).
[12] R. Al-Msie’deen, A. Seriai, M. Huchard, C.
Urtado, and S. Vauttier, “Mining features from the
object-oriented source code of software variants
by combining lexical and structural similarity,”
IRI, 586 – 593, (2013).
[13] P. Warintarawej, M. Huchard, M. Lafourcade, A.
Laurent, and P. Pompidor, “Software
understanding: Automatic classification of
software identifiers,” Intell. Data Anal., 19(4): 761
– 778, (2015).
[14] A. Wiese, V. Ho, and E. Hill, “A comparison of
stemmers on source code identifiers for software
search,” ICSM, 496 – 499, (2011).
[15] R. Al-Msie’deen, A. Seriai, M. Huchard, C.
Urtado, S. Vauttier, and H. E. Salman, “Mining
features from the object-oriented source code of
a collection of software variants using formal
concept analysis and latent semantic indexing,”
SEKE, 244–249, (2013).
[16] M. V. Couto, M. T. Valente, and E. Figueiredo,
“Extracting software product lines: A case study
using conditional compilation,” CSMR, 191 –
200, (2011).
[17] J. Emerson, N. Churcher, and A. Cockburn, “Tag
clouds for software and information
visualisation,” ACM SIGCHI_NZ, 1 – 4, (2013).
[18] R. Al-Msie’deen, M. Huchard, A. Seriai, C.
Urtado, and S. Vauttier, "Reverse engineering
feature models from software configurations
using formal concept analysis," Proceedings of
the Eleventh International Conference on
Concept Lattices and Their Applications, 95 –
106, (2014).](https://image.slidesharecdn.com/1803-230126152422-7d6a2942/85/Automatic-Labeling-of-the-Object-oriented-Source-Code-The-Lotus-Approach-4-320.jpg)
Automatic Labeling of the Object-oriented Source Code: The Lotus Approach
- 1. Sci.Int.(Lahore),30(1),45-48,2018 ISSN 1013-5316;CODEN: SINTE 8 45 January-February AUTOMATIC LABELING OF THE OBJECT-ORIENTED SOURCE CODE: THE LOTUS APPROACH Ra'Fat Al-Msie'deen Department of IT, Faculty of Information Technology, Mutah University, P.O. Box 7, Mutah 61710, Karak, Jordan Email: rafatalmsiedeen@mutah.edu.jo ABSTRACT: Most of open-source software systems become available on the internet today. Thus, we need automatic methods to label software code. Software code can be labeled with a set of keywords. These keywords in this paper referred as software labels. The goal of this paper is to provide a quick view of the software code vocabulary. This paper proposes an automatic approach to document the object-oriented software by labeling its code. The approach exploits all software identifiers to label software code. The paper presents the results of study conducted on the ArgoUML and drawing shapes case studies. Results showed that all code labels were correctly identified. Keywords: Software engineering, Software comprehension, Reverse engineering, Software visualization, Code Label. 1. INTRODUCTION Program understanding is the main activity for software maintenance and evolution [1]. The main idea of this paper is to reduce the amount of information mined from the software source code to provide a quick overview of the software code vocabularies. This paper proposes a novel approach called Lotus to automatically provide labels for software code. Name of the approach inspired by the Lotus flower. Lotus stands for automatic Labeling of the ObjecT-oriented SoUrce code. Several companies are facing problems with legacy software systems such as: software understanding and maintenance. The reason behind these problems is the absence of software documentation [2]. The quick development of software engineering approaches and the increasing of software system complexity has led to a large production of textual information contained in software artifacts (e.g., software source code and design documents). As a result, numerous papers investigated the analysis of textual information contained in the software artifacts to support software engineering activities [3] such as: feature location [4] and software documentation [5]. To extract labels from the software source code, the Lotus approach relies on the software identifiers (i.e., package, class, attribute and method). Basically, the most important words are included in the package, class, method, and attribute names [6]. The extracted labels can give a hint to the software developers about software code vocabularies. The software code labeling process aims to analyze the legacy software to determine its main keywords, where the huge amounts of code need to be understood. To help a human expert to document the legacy software, the Lotus approach generates a set of labels represent the software vocabularies. This paper proposes an automatic approach which extracts labels from software using its source code. Compared with the existing work that labeling software source code (cf. section 2), the novelty of the Lotus approach is that it exploits all software identifiers to generate software labels in an efficient way. Lotus approach identifies the software code using static code parser. Then, the approach identifies all software identifiers. Then, Lotus splits the name of identifier into a set of keywords based on the CamelCase methods [7]. Then, Lotus returns each keyword to its stem based on the WordNet1 . Finally, Lotus generates software labels. A 1 WordNet: https://wordnet.princeton.edu/ sample execution of the Lotus approach is displayed on Figure 1. Figure 1. Running the Lotus approach on a simple example. Lotus approach is detailed in the remainder of this paper as follows. Section 2 discusses the related work. Section 3 shows an overview of Lotus approach. Section 4 presents the source code labeling process step by step. Section 5 describes the experimentation, while section 6 concludes and provides perspectives for this work. 2. RELATED WORK This section presents the related work and provides a concise summary of the different approaches. Kuhn [8] presents a lexical approach that uses the log- likelihood ratios of word frequencies to automatically provide labels for software components. AL-Msie'Deen [5] developed a tool called Vsound to visualize the software code and its dependencies. Vsound relies on software identifiers to visualize and document the
- 2. 46 ISSN 1013-5316;CODEN: SINTE 8 Sci.Int.(Lahore),30(1),45-48,2018 January-February software code. While Lotus approach extracts labels from software identifiers. AL-Msie'Deen et al. [9] proposed an approach called REVPLINE to retrieve labels for the mined features from the source code of software family based on the use-case diagrams. REVPLINE gives as output for each feature implementation, a label based on the use-case label [10]. Lucia et al. [6] suggest a method for source code labeling, based on information retrieval techniques, to identify relevant words in the source code. They applied numerous information retrieval methods [1] to extract terms from the names of software classes. Their approach does not consider the names of packages, attributes and methods. Lotus approach considers all granularity levels of software code. Kuhn et al. [11] suggest a method to cluster software classes that use similar vocabulary together. Then, they use latent semantic indexing to automatically label clusters with their most relevant terms. Lotus approach aims to document the software code as a collection of labels. Most existing approaches are designed to extract labels to name software features, components or clusters. In the literature, there is no work identifies code labels at all granularity levels. Also, many approaches manually extract software labels. Conversely, Lotus is designed to automatically extract software labels based on all software identifiers. 3. APPROACH OVERVIEW This section provides an overview of code labeling process and describes the example that illustrates the remaining of the paper. The main goal of Lotus approach is to understand the software code vocabularies. The Lotus approach aims to provide the software developers with software labels. The complex software system involves a huge amount of source code (i.e., lines of code), so there are needs to deal with all granularity levels of source code in an abstract way. Figure 2. The source code labeling process. The source code labeling process takes the software's source code as its inputs and creates code labels as its outputs. The goal of these labels is to provide us with a quick overview of software vocabularies. The software identifier names play a significant role in software understanding, particularly when the software documents are missing, or when the software code is very complex such that the identifier names would tell more about code to the software developers [6]. Figure 2 shows the source code labeling process in general. Lotus approach extracts software code. Then, the approach identifies software identifiers. Then, the approach splits the software identifiers into keywords based on the CamelCase method. Then, the Lotus returns each keyword to the word root via WordNet. Finally, Louts builds software labels. As an illustrative example, Lotus considers the drawing shapes software [12]. The software allows user to draw three types of shapes which are lines, rectangles and ovals2 . This software used to better explain the code labeling process. 4. SOURCE CODE LABELING PROCESS This section describes the source code labeling process step by step. Lotus approach identifies the source code labels in five steps as detailed in the following. 4.1 Extracting source code The Lotus approach only uses the software source code as input of the labeling process and thus Lotus does not know the code labels in advance. The first step of the labeling process aims to extract the software source code. The Lotus approach extracts all essential information from source code such as: software identifiers and code dependencies [5]. The static code parser3 takes software code to generate the code file as output. The code file contains software identifiers. The Lotus source code parser is written completely in Java. 4.2 Identifying software identifiers To comprehend the source code of legacy software systems, it's important to work with all granularity levels of the source code. Lotus approach provides us with four kinds of documents (aka identifiers file). Each document contains a set of identifier names. This step generates four documents which are package, class, attribute and method document. Table 1. Examples of software identifiers from drawing shapes software. Package Names Class Names Drawing MyLine Drawing.Shapes MyOval Drawing.Shapes.coreElements MyRectangle Drawing.Shapes.coreFrame PaintJPanel DrawingShapes Examples of software identifiers are shown in Table 1. The package document contains the names of all software packages. The class document contains the names of all software classes and so on. 4.3 Splitting the identifiers The identifier names are used to define the main entities of the software system (e.g., package and class). The names of the identifiers represent a set of characters based on the rules of programming languages [13]. Lotus relies on the CamelCase technique as an identifier splitting algorithm. CamelCase method is a simple and generally used method for identifier splitting algorithms [7] 2 Drawing shapes: https://sites.google.com/site/ralmsideen/tools 3 Static code parser: https://sites.google.com/site/ralmsideen/tools
- 3. Sci.Int.(Lahore),30(1),45-48,2018 ISSN 1013-5316;CODEN: SINTE 8 47 January-February and the rules of splitting are widely based on CamelCase agreement. Lotus splits identifier names into a set of keywords based on the camel-case syntax. For example, the identifier DrawingShapes is split into two words: drawing and Shapes. Table 2. Examples of the camel-case identifier splitting algorithm. Identifier Name Token/word Word-1 Word-2 Word-3 shapesArrayList shapes array list setCurrentColor set current color colorJButton color j button createUserInterface create user interface Examples of camel-case splitting algorithm are shown in Table 2. 4.4 Stemming identifier keywords Stemming is the procedure of stripping affixes (i.e., prefixes and suffixes) from words to form a stem [14]. Stemming is often applied to words in information retrieval methods (e.g., latent semantic indexing [12]). The stemming is used in Lotus to replace English words with their root or stem [15]. Lotus returns each keyword to its root or word stem. For instance, if we have the following keyword "drawing", the root of this keyword will be "draw". After the stemming was performed via WordNet the stems of keywords were stored in the terms file to generate the software labels. Table 3. A sample of the word stems from drawing shapes software. Identifier word Word stem Pressed Press Drawing Draw Performed Perform Dragged Drag Table 3 shows a sample of the word stems (i.e., terms) from drawing shapes software. 4.5 Building the software labels In this step, Lotus approach builds software labels to show a quick view of the used vocabularies in the software code. Figure 3 shows the mined labels from drawing shapes software. Figure 3. The extracted labels from drawing shapes software. In Lotus approach, the extracted labels called label map. All information about source code vocabulary given in the extracted label map. Thus, all code vocabularies can easily be acquired when needed via Lotus approach. 5. EXPERIMENTATION This section presents the experiment that conducted to validate the Lotus approach. Firstly, this section presents the ArgoUML case study. Then, it presents the source code labeling outcomes and, at last, it presents the threats to validity of the Lotus approach. In addition to the toy example used in this paper (i.e., drawing shapes software), the Lotus approach has been tested on the ArgoUML4 software system. ArgoUML is a widely used open source tool for UML modeling tool [15] and well known [16]. ArgoUML software represents a large case study (i.e., 120,348 lines of code). Figure 4. The extracted package label map from ArgoUML. Figure 4 shows the extracted package label map from ArgoUML software using Lotus prototype5 . The algorithm execution time is equal to 50860 ms. The identified labels (aka topics) are very helpful to document software vocabularies. In Figure 4, the package labels presented in an alphabetical order and without repetition. Figure 5 shows the extracted class label map from ArgoUML software. The software developer can easily obtain all code labels using Lotus approach. For a lack of studies that are evaluating the extracted labels from software code, there was difficulty in evaluating the Lotus approach. Figure 5. The extracted class label map from ArgoUML. The threat to the validity of Lotus is that camel-case syntax may be not reliable in all cases to split software identifiers. Also, Lotus considers only the Java software, and this limits the Lotus implementation ability to deal only with Java language. Also, if the label map comes up with labels such as: m, l, get or in this does not tell a human expert 4 ArgoUML: http://argouml.tigris.org/ 5 Lotus prototype: https://sites.google.com/site/ralmsideen/tools
- 4. 48 ISSN 1013-5316;CODEN: SINTE 8 Sci.Int.(Lahore),30(1),45-48,2018 January-February much about the software system. In this case, the label map was totally useless, since the developer knowledge is missing [11]. Table 4: Summary of Lotus approach. Objectives Programmed method Code labeling √ Automatic √ Code understanding √ Input Code documentation √ Software packages √ Code visualization √ Software classes √ Tool support Software attributes √ WordNet √ Software methods √ Evaluation Technique Evaluation x Ad hoc algorithm √ Case study Camel-case algorithm √ Drawing shapes √ Splitting method ArgoUML √ Camel-case √ Output Package label map √ Class label map √ Attribute label map √ Method label map √ Table 4 summarizes Lotus approach where it presents the objectives, programmed method, input, tool support, evaluation, technique, case study, splitting method and output. 6. CONCLUSION AND FUTURE WORK This paper proposed an approach to extract code labels. The goal of this paper is to document code vocabularies. The novelty of this paper is that it exploits all software identifiers to document software vocabularies. Lotus has implemented on several case studies. Results showed that all label maps were identified. Regarding future work, Lotus plans to use code comments in the code labeling process. Also, Lotus approach plans to exploit the tag cloud visualization technique [17] to display label frequency across software code. In addition, Lotus plans to study the evolution of software labels in a set of software variants using formal concept analysis [18]. 7. REFERENCES [1] A. D. Lucia, M. D. Penta, R. Oliveto, A. Panichella, and S. Panichella, “Labeling source code with information retrieval methods: an empirical study,” Empirical Software Engineering, 19(5): 1383 – 1420, (2014). [2] S. Souza, N. Anquetil, and K. Oliveira. “A study of the documentation essential to software maintenance,” SIGDOC '05, 68 – 75 (2005). [3] R. Al-Msie’deen, A. Seriai, M. Huchard, C. Urtado, S. Vauttier, and H. E. Salman, “Feature location in a collection of software product variants using formal concept analysis,” ICSR, 302 – 307, (2013). [4] R. Al-Msie’deen, “Reverse engineering feature models from software variants to build software product lines: REVPLINE approach,” Ph.D. dissertation, Montpellier 2 University, France, (2014). [5] R. Al-Msie’deen, “Visualizing object-oriented software for understanding and documentation,” IJCSIS, 13(5): 18 – 27, (2015). [6] A. D. Lucia, M. D. Penta, R. Oliveto, A. Panichella, and S. Panichella, “Using IR methods for labeling source code artifacts: Is it worthwhile?” ICPC, 193 – 202, (2012). [7] B. Dit, L. Guerrouj, D. Poshyvanyk, and G. Antoniol, “Can better identifier splitting techniques help feature location?”, ICPC, 11 – 20, (2011). [8] A. Kuhn, “Automatic labeling of software components and their evolution using log- likelihood ratio of word frequencies in source code,” MSR, 175 – 178, (2009). [9] R. Al-Msie’deen, A. Seriai, M. Huchard, C. Urtado, and S. Vauttier, “Documenting the mined feature implementations from the object-oriented source code of a collection of software product variants,” SEKE, 138 – 143, (2014). [10] R. Al-Msie’deen, M. Huchard, A. Seriai, C. Urtado, and S. Vauttier, “Automatic documentation of [mined] feature implementations from source code elements and use-case diagrams with the REVPLINE approach,” IJSEKE, 24(10): 1413 – 1438, (2014). [11] A. Kuhn, S. Ducasse, and T. Gîrba, “Semantic clustering: Identifying topics in source code,” Information & Software Technology, 49(3): 230 – 243, (2007). [12] R. Al-Msie’deen, A. Seriai, M. Huchard, C. Urtado, and S. Vauttier, “Mining features from the object-oriented source code of software variants by combining lexical and structural similarity,” IRI, 586 – 593, (2013). [13] P. Warintarawej, M. Huchard, M. Lafourcade, A. Laurent, and P. Pompidor, “Software understanding: Automatic classification of software identifiers,” Intell. Data Anal., 19(4): 761 – 778, (2015). [14] A. Wiese, V. Ho, and E. Hill, “A comparison of stemmers on source code identifiers for software search,” ICSM, 496 – 499, (2011). [15] R. Al-Msie’deen, A. Seriai, M. Huchard, C. Urtado, S. Vauttier, and H. E. Salman, “Mining features from the object-oriented source code of a collection of software variants using formal concept analysis and latent semantic indexing,” SEKE, 244–249, (2013). [16] M. V. Couto, M. T. Valente, and E. Figueiredo, “Extracting software product lines: A case study using conditional compilation,” CSMR, 191 – 200, (2011). [17] J. Emerson, N. Churcher, and A. Cockburn, “Tag clouds for software and information visualisation,” ACM SIGCHI_NZ, 1 – 4, (2013). [18] R. Al-Msie’deen, M. Huchard, A. Seriai, C. Urtado, and S. Vauttier, "Reverse engineering feature models from software configurations using formal concept analysis," Proceedings of the Eleventh International Conference on Concept Lattices and Their Applications, 95 – 106, (2014).