Major change
[u/philim/db2osl_thesis.git] / program_code.tex
... / ...
1\section{Code style}
3TODO: Conventions, ex.: iterators\\
4As the final system hopefully will have a long lifetime cycle
5and will be used and refined by many people, high code quality was an important aim.
6Beyond architectural issues this also involves cleanness on the lower level,
7like the design of classes and the implementation of methods.
8Common software development principles were followed and
9the unfamiliar reader was constantly taken into account
10to yield clean, readable and extensible code.
12%Scarce, informative comments were inserted at places of higher complexity and to
13%expose logical subdivisions, but care was taken to use ``speaking code'' in favor
14%of rampant comments.
15%Complex, misleading and hard-to-use interfaces were avoided wherever possible.
16%External software libraries employed were chosen to be stable, specific,
17%well structured, publicly available and ideally in wide use.
21Comments were used at places ambiguities or misinterpretations could arise,
22yet care was taken to face such problems at their roots and solve them
23wherever possible instead of just effacing the ambiguity with comments.
24This approach is further explained in section \fullref{speaking} and
25rendered many uses of comments unnecessary.
27In fact, the number of (plain, e.g. non-\name{Javadoc}) comments was
28consciously minimized, to enforce speaking code and avoid redundancy.
29An exception from this was the highlighting of subdivisions.
30In class and method implementations, comments like
31\codepar{//********************** Constructors **********************\textbackslash\textbackslash}
33were deliberately used to ease navigation inside source files,
34but also to enhance readability: parts of method
35implementations, for example, were optically separated this way.
36Another alternative would have been to use separate methods for these code
37pieces, and thereby sticking strictly to the so-called
38``Composed Method Pattern'' \cite{composed},
39as was done in other cases.
40However, sticking to this pattern too rigidly would have introduced additional
41artifacts with either long or non-speaking names,
42would have interrupted the reading flow and also would have increased complexity,
43because these methods would have been callable at least from everywhere
44in the source file.
45Consequently, having longer methods at some places that are optically separated
46into smaller units that are in fact independent from each other was considered
47an elegant solution, although, surprisingly, this technique does not seem to be
48proposed that often in the literature.
50Wherever possible, the appropriate \name{Javadoc} comments were used in favor of
51plain comments, for example to specify parameters, return types, exceptions
52and links to other parts of the documentation.
53This proved even more useful due to the fact that \name{Doxygen} supports all
54of the used \name{Javadoc} comments \cite{doxygen} (but not vice versa \cite{javadoc}).
56\subsection{``Speaking code''}
58As mentioned in section \fullref{comments}, the code was tried to be designed to
59``speak for itself'' as much as possible instead of making its readers depend on
60comments that provide an understanding.
61In doing so, besides reducing code size due to the missing comments,
62clean code amenable to unfamiliar readers and unpredictable changes was enforced.
63This is especially important since, as described in section \fullref{arch},
64\myprog{} was designed to not only be a standalone program but also offer
65components suitable for reusability.
67TODO: understandability <- code size
69The following topics were identified to be addressed to get what can be
70conceived as ``speaking code'':
72 \item Meaningful typing
73 \item Method names
74 \item Variable names
75 \item Intuitive control flow
76 \item Limited nesting
77 \item Usage of well-known structures
80~\\The rest of this section describes these topics in some detail.
81Besides, an intuitive architecture and suitable, well-designed libraries
82also contribute to the clarity of the code.
84\subsubsection{Meaningful typing}
85Meaningful typing includes the direct mapping of entities of the modeled world
86to code entities \cite{str4} as well as an expressive naming scheme
87for the obtained types.
88Furthermore, inheritance should be used to express commonalities, to avoid
89code duplication and to separate implementations from interfaces \cite{str4}.
91All real-world artifacts to be modeled like database schemata, tables, table schemata.
92columns, keys and OBDA specifications with their certain map types were directly
93translated into classes having simple predicting names like \code{Table},
94\code{TableSchema} and \code{Key}.
95Package affiliation provided the correct context to unambiguously understand these names.
97\subsubsection{Method names}
98Assigning expressive names to methods is a substantially important part of
99producing speaking code, since methods encapsulate operation and as such
100are important ``building blocks'' for other methods \cite{str4} and ultimately
101the whole program.
102Furthermore, method names often occur in interfaces and therefore are not limited
103to a local scope, and neither are easily changeable without affecting callers
106Ultimately, care was taken that method names reflect all important aspects
107of the respective method's behavior.
108Consider the following method from \file{}:
109\codepar{public static void promptAbortRetrieveDBSchemaAndWait\\
110 \ind(final FutureTask<DBSchema> retriever) throws SQLException}
112It could have been called \code{promptAbortRetrieveDBSchema} only, with the
113waiting mentioned in a comment.
114However, the waiting (blocking) is such an important part of its behavior, that this
115was considered not enough, so the waiting was included in the function name.
116Since the method is called at one place only, the lengthening of the method
117name by 7 characters or about 26 \% is really not a problem.
119\subsubsection{Variable names}
120To keep implementation code readable, care was taken to name variables
121meaningful yet concise. If this was not possible, expressiveness was preferred
122over conciseness.
124For example, in the implementation of the database schema retrieval,
125variables containing data directly obtained from querying the database
126and thus being subject to further processing was consequently prefixed
127with ``\code{recvd}'', although in most cases this technically would not have
128been necessary.
130\subsubsection{Intuitive control flow}
131To consequently stick to the maxim of speaking code and further increase readability,
132control flow was tried to kept intuitive.
133\code{do-while} loops, for example, are unintuitive: they complicate matters due to
134the additional, unconditional, loop their reader has to keep in mind.
135Even worse, \name{Java}'s Syntax delays the occurrence of their most important
136control statement -- the loop condition -- till after the loop body.
137Usually, \code{do-while} loops can be circumvented by properly setting variables
138influencing the loop condition immediately before the loop and using a \code{while}
140Consequently, \code{do-while} loops were omitted -- the code of \myprog{} does not
141contain a single \code{do-while} loop.
142TODO: references
144Another counterproductive technique is the avoidance of the advanced loop control
145statements \code{break}, \code{continue} and \code{return} and the sole direction
146of a loop's control flow with its loop condition, often drawing on additional
147boolean variables like \code{loopDone} or \code{loopContinued}.
148This approach is an essential part of the ``structured programming (paradigm)''
149\cite{struc} and its purpose is to enforce that a loop is always
150left regularly, by unsuccessfully checking the loop condition, which shall ease
151code verification \cite{struc}.
152A related topic is the general avoidance of the \code{return} statement (except at
153the end of a method) for similar considerations \cite{struc}.
154However, both are not needed \cite{clean} and, as always, the introduction
155of artificial technical constructs impairs readability and the ability of the code
156to ``speak for itself''.
158Consequently, control flow was not distorted for technical considerations
159and care was taken to yield straight-forward loops, utilizing advanced control
160statements to be concise and intuitive and cleverly designed methods that benefit
161from well-placed \code{return} statements.
163\subsubsection{Limited nesting}
164A topic related to intuitive control flow is limited code nesting.
165Most introductions of new nesting levels greatly increase complexity,
166since the associated conditions for the respective code to be reached
167combine with the previous ones in often inscrutable ways.
168Besides being aware of the execution condition for the code he is currently
169reading, the reader is forced to either remember the sub-conditions introduced
170with each nesting level, as well as the current nesting level,
171or to jump back to the introduction of one or more nestings to figure out the
172relevant execution condition again.
174Naturally, such code is far from being readable and expressive.
175Thus, overly deep nesting was avoided by rearranging code or using control
176statements like \code{return} in favor of opening a new \code{if} block.
177The deepest and most complicated nesting in \myprog{} has level $5$
178(with normal, non-nested method code having level $0$), with
179one of these nestings being dedicated to a big enclosing \code{while} loop,
180one to a \code{try-catch} block and the remaining three to \code{if} blocks
181with no \code{else} parts and trivial one-expression conditions.
182Additionally, in this case all of the nesting blocks only contained
183a few lines of code, making the whole construction easily fit on one screen,
184so this was considered all right.
185At a few other places there occurs similar, less complicated, nesting up to level $5$.
186%These were similar to the above but with two enclosing loops and one or
187%even two \code{try-catch} blocks.
188TODO: references
190\subsubsection{Usage of well-known structures}
191Great benefit can be taken from constructs familiar to programmers
192regarding expressiveness.
193Surely, implementations based on such well-known constructs and patterns
194are much more likely to be instantly understood by programmers and therefore
195have a much higher ability of ``speaking for themselves''.
197Examples in \myprog{} are the (extensively used) iterator concept,
198const correctness (see paragraph ``\nameref{const}''
199in section \fullref{code_classes}), exceptions, predicates \cite{str4},
200run-time type information \cite{str4}, helper functions \cite{str4}
201and well-known interfaces from the \name{Java} API like \code{Set} or
202\code{Collection}, as well as common \name{Java} constructs, like
203classes performing a single action (e.g. \code{OSLSpecPrinter}), and
204naming schemes, like \code{get...}/\code{set...}/\code{is...}.
206\subsection{Robustness against incorrect use}
207Care was taken to produce code that is geared to incorrect use, making it
208suitable for the expected environment of sporadic updates by unfamiliar and
209potentially even unpracticed programmers, who besides have their emphasis
210on the concepts of bootstrapping rather than details of the present code anyway.
211In fact, carefully avoiding the introduction of technical artifacts to mind,
212preventing programmers from focusing on the actual program logic,
213is an important principle of writing clean code \cite{str4}.
215In modern object-oriented programming languages, of course the main instruments
216for achieving this are the type system and exceptions.
217In particular, static type information should be used to reflect data
218abstraction and the ``kind'' of data, an object reflects,
219while dynamic type information should only be used implicitly,
220through dynamically dispatching method invocations \cite{str3}.
221Exceptions on the other hand should be used at any place related to errors
222and error handling, separating error handling noticeably from other code and
223enforcing the treatment of errors \cite{str4}, preventing the programmer from using
224corrupted information in many cases.
226An example of both mechanisms, static type information and exceptions, acting
227in combination, while cleanly fitting into the context of dynamic dispatching,
228are the following methods from \file{}:
229\codepar{public Boolean isNonNull()\\public Boolean isUnique()}
231Their return type is the \name{Java} class \code{Boolean}, not the plain type
232\code{boolean}, because the information they return is not always known.
233In an early stage of the program, they returned \code{boolean} and were
234accompanied by two methods
235\code{public boolean knownIsNonNull()} and \code{public boolean knownIsUnique()},
236telling the caller whether the respective information was known and thus the
237value returned by \code{isNonNull()} or \code{isUnique()}, respectively,
238was reliable.
240They were then changed to return the \name{Java} class \code{Boolean} and to return
241null pointers in case the respective information is not known.
242This eliminated any possibility of using unreliable data in favor of generating
243exceptions instead, in this case a \code{NullPointerException}, which is thrown
244automatically by the \name{Java} Runtime Environment \cite{java} if the programmer
245forgets the null check and tries to get a definite value from one of these methods
246when the correct value currently is not known.
248Comparing two unknown values -- thus, two null pointers --
249also yields the desired result, \code{true}, since the change,
250even when the programmer forgets that he deals with objects.
251However, when comparing two return values of one of the methods in general
252-- as opposed to comparing one such return value against a constant --,
253errors could occur if the programmer mistakenly writes \code{col1.isUnique() == col2.isUnique()}
254instead of \code{col1.isUnique().booleanValue() == col2.isUnique().booleanValue()}.
255In this case, since the two \code{Boolean} objects are compared for identity \cite{java},
256the former comparison can return \code{false}, even when the two boolean values are in fact
257the same.
258However, since this case was considered much less common than cases in which the other
259solution could make programmers making mistakes produce undetected errors, it was preferred.
261TODO: summary
263\subsection{Use of classes}
265Following the object-oriented programming paradigm \cite{obj}, classes were heavily used
266to abstract from implementation details and to yield intuitively usable objects with
267a set of useful operations.
269\subsubsection{Identification of classes}
270To identify potential classes, entities from the problem domain were -- if reasonable --
271directly represented as \name{Java} classes.
272The approach of choosing ``the program that most directly models the aspects of the
273real world that we are interested in'' to yield clean code,
274as described and recommended by Stroustrup \cite{str3}, proved to be extremely useful
275and effective.
276As a consequence, the code declares classes like \code{Column}, \code{ColumnSet},
277\code{ForeignKey}, \code{Table}, \code{TableSchema} and \code{SQLType}.
278As described in section \fullref{speaking}, class names were chosen to be concise
279but nevertheless expressive.
280\name{Java} packages were used to help attain this aim,
281which is why the previously mentioned class names are unambiguous.
282For details about package use, see section \fullref{code_packages}.
284Care was taken not to introduce unnecessary classes, thereby complicating
285code structure and increasing the number of source files and program entities.
286Especially artificial classes, having little or no reference to real-world
287objects, could most often be avoided.
288On the other hand of course, it usually is not the cleanest solution
289to avoid such artificial classes entirely.
291Section \fullref{hierarchies} describes how the classes of \myprog{} are organized
292into class hierarchies.
294\subsubsection{Const correctness}
296Specifying in the code which objects may be altered and which shall remain constant,
297thus allowing for additional static checks preventing undesired modifications,
298is commonly referred to as ``const correctness'' TODO.
299TODO: powerful, preventing errors, clarity
301Unfortunately, \name{Java} lacks a keyword like \name{C++}'s \code{const},
302making it harder to achieve const correctness \cite{final}.
303It only specifies the similar keyword \code{final}, which is much less expressive and
304doesn't allow for a similarly effective error prevention \cite{final}.
305In particular, because \code{final} is not part of an object's type information,
306it is not possible to declare methods that return read-only objects \cite{final} --
307placing a \code{final} before the method's return type would declare the
308method \code{final} \cite{java}.
309Similarly, there is no way to express that a method must not change
310the state of its object parameters. A method like \code{public f(final Object obj)}
311is only liable to not assigning a new value to its parameter object \code{obj} \cite{java}
312(which, if allowed, wouldn't affect the caller anyway \cite{java}).
313Methods changing its state, on the other hand,
314are allowed to be called on \code{obj} without restrictions.
316Several possibilities were considered to address this problem:
318 \item Not implementing const correctness, but stating the access rules in
319 comments only
320 \item Not implementing const correctness, but giving the methods which modify
321 object states special names like
322 \code{setName\textendash\textendash USE\_WITH\_CARE}
323 \item Implementing const correctness by delegating changes of objects
324 to special ``editor'' objects to be
325 obtained when an object shall be modified
326 \item Implementing const correctness by deriving classes offering
327 the modifying methods from read-only classes
330Not implementing const correctness at all of course would have been the simplest
331possibility, producing the shortest and most readable code, but since
332incautious manipulation of objects would possibly have introduced subtle,
333hard-to-spot errors which in many cases would have occurred under additional
334conditions only and at other places, for example when inserting a \code{Column}
335into a \code{ColumnSet}, this method was not seriously considered.
337Not implementing const correctness but using intentionally angular,
338conspicuous names also was not considered seriously,
339since it would have cluttered the code for the only sake of hopefully warning
340programmers of possible errors -- and not attempting to avoid them technically.
342So the introduction of new classes was considered the most effective and cleanest
343solution, either in the form of ``editor'' classes or derived classes offering the
344modifying methods directly. Again -- as during the identification of classes --,
345the most direct solution was considered the best, so the latter form of introducing
346additional classes was chosen and classes like \code{ReadableColumn},
347\code{ReadableColumnSet} et cetera were introduced which offer only the read-only
348functionality and usually occur in interfaces.
349Their counterparts including modifying methods also were derived from them and the
350implications of modifications were explained in their documentation, while the
351issue and the approach as such were also mentioned in the documentation of the
352\code{Readable...} classes.
353The \code{Readable...} classes can be converted to their fully-functional
354counterparts via downcasting (only), thereby giving a strong hint to
355programmers that the resulting objects are to be used with care.
357\subsubsection{Java interfaces}
359In \name{Java} programming, it is quiet common and often recommended \cite{gof}
360that every class has at least one \code{interface} it \code{implements},
361specifying the operations the class provides.
362If no obvious \code{interface} exists for a class or the desired
363interface name is already given to some other entity,
364the interface is often given names like \code{ITableSchema}
365or \code{TableSchemaInterface}.
367However, for a special purpose program with a relatively fixed set of classes
368mostly representing real-world artifacts from the problem domain,
369this approach was considered overly cluttering, introducing artificial
370code entities for no benefit.
371In particular, as explained in section \fullref{fine}, all program classes either
372are standing alone or belong to a class hierarchy derived from at least one
374So, except from the standalone classes, an interface existed anyway, either
375``naturally'' (as in the case of \code{Key}, for example) or because of
376the chosen way to implement const correctness.
377In some cases, these were interfaces declared in the program code, while
378in some cases, \name{Java} interfaces like \code{Set} were implemented
379(an obvious choice, of course, for \code{ColumnSet}).
380Introducing artificial interfaces for the standalone classes was considered
381unnecessary at least, if not messy.
383\subsection{Use of packages}
385As mentioned in section \fullref{code_classes}, class names were chosen to be
386concise but nevertheless expressive.
387This only was possible through the use of \name{Java} \code{package}s,
388which also helped structure the program.
390For the current, relatively limited, extent of the program which currently
391comprises $45$ (\code{public}) classes, a flat package structure was
392considered ideal, because it is simple and doesn't stash source files deep
393in subdirectories (in \name{Java}, the directory structure of the source tree
394is required to reflect the package structure \cite{java}).
395Because also every class belongs to a package,
396each source file is to be found exactly one directory below the root
397program source directory, which in many cases eases their handling.
399For the description of the packages, their interaction and considerations on
400their structuring, see section \fullref{coarse}.
401For a detailed package description, refer to Appendix TODO.
403Each package is documented in the source code also, namely in a file
404\file{} residing in the respective package directory.
405This is a common scheme supported by the \name{Eclipse} IDE as well as the
406documentation generation systems \name{Javadoc} and \name{Doxygen}
407(all of which were used in the creation of the program,
408as described in section \fullref{tools}).