Control Structures Programs have 4 basic control structures

Control Structures • Programs have 4 basic control structures: – – Sequential Selection Repetition Unconditional Branching • we can break unconditional branches into those that pass parameters (function calls) and those that do not, and whether they return to the current location (function calls) or not (GO TO types) • For purposes of local scope and easier syntax, we can add blocks or compound statements – ALGOL 60 - first language to introduce blocks using beginend statements (also used in Pascal, Ada, Modula, etc) – C, C++, Java, PHP, Java. Script, Perl - uses { } as delimiters – Lisp uses ( ) • Control statements are essential for any programming language

Selection Statements • Gives a program the ability to choose which instruction(s) to next execute based on conditions • Types: – 1 -Way selection (if statement) – 2 -Way selection (if-else statement) – 3 -Way selection (FORTRAN had a peculiar arithmetic-based selection statement) – multiple selection (or n-way selection, switch/case) • Design issues: – what is the form and type of expression that controls the selection? (C allows 0/non-0, Java/C# allow only boolean) – how are clauses specified if at all? – if nesting is allowed, how is it specified and implemented?

One-way and Two-way Selections • One-way: if without else – if condition is true then execute next statement, otherwise skip over it • FORTRAN’s IF statement was a One-Way (no Else clause, no nesting) • if then clause has more than 1, instruction, we alter the semantics of the statement, to what is in essence an “else-goto” rather than, an “if-then” If (. NOT. Condition) GOTO 20 I=1 J=2 20 Continue • nearly every language since has allowed two-way selections • ALGOL 60 introduced the first true two-way selection – terms “then-clause” & “else-clause” introduced – clauses are expected to be a single instruction to easily detect the end of each clause, otherwise, must use blocks – ALGOL 60 allowed more than one entry into the two-way selection clauses! – in languages where the “then” is omitted, the condition must be placed in ( ) to denote its end

Blocks • ALGOL introduced the block structure in part to permit multiple instructions being part of then or else clauses – these required delimiters – begin. . end, { }, ( ) • in Algol 60: • If (Boolean expression) then begin statement 1; . . . statement n end; – Perl goes one step further and requires that all clauses be placed into blocks (even if the clause is a single instruction) – Ada and FORTRAN 95 have explicit end of blocks but not explicit beginnings by using end if or end for statements

Nesting • Nested If-Then-Else statements can lead to ambiguity if there is a miss-match between conditions and else clauses if(sum = = 0) if (count = = 0) result = 0; else result = 1; Which condition does the else go with? Is result = 1 if sum != 0 or if sum = = 0 and count != 0? • In ALGOL 60, nested if-then-else clauses must use an explicit end statement • this is also the case for FORTRAN 77/90/95, Modula-2 and Ada • In Ruby, all clauses must have explicit end statements (even if there is only a single statement in the clause) • In C/C++/Java/C# and Pascal, the compiler rule matches mismatched elses with the last unmatched condition – the rule can be overridden using blocks if sum = = 0 then if count = = 0 then result = 0 else result = 1 end if sum = = 0 then if count = = 0 then result = 0 end else result = 1 end

Multiple Selection Constructs • FORTRAN offers a 3 -way selection – IF (expression) N 1, N 2, N 3 • if expression < 0 then goto N 1, if = 0 goto N 2, if > 0 goto N 3 – this was FORTRAN I-IVs only IF statement! • ALGOL-W introduced the case statement – Pascal, Modula and Ada, FORTRAN 90/95 all use this – C/C++/Java/C# use the switch statement • the main difference is that after a condition is found true in the switch statement, the next condition is tested such that the switch statement is not exited until all cases have been tested, to avoid this, you must use break statements • however, to fix this oddity, in C# you must end each case with a break or goto – Lisp uses the COND statement – all of these allow for a default if none of the cases is selected • default in C-languages, else in Pascal, when others in Ada • Perl/Python don’t have a multi-selection statement at all!

Nested If-Else Constructs • Cond in Lisp is really a nested if-then-else function – Lisp also provides a “default” by using T as the final condition • Other languages have provided a specific if-then-else nested construct for convenience – Ada uses elsif if the intention is to do “else if” so that you do not have to explicitly end each if statement with an end if – Python uses elif so that you don’t have to continue to indent Ada without elsif: if Count < 10 then Bag 1 : = True; else if Count < 100 then Bag 2 : = True; else if Count < 1000 then Bag 3 : = True; end if; Ada with elsif: if Count < 10 then Bag 1 : = True; elsif Count < 100 then Bag 2 : = True; elsif Count < 1000 then Bag 3 : = True; end if;

Repetition Statements • Every language has included some form of repetition, either counter-controlled (early FORTRAN) or logically -controlled, or both • Issues: – – how is repetition controlled? is testing before or after the loop body? (pre vs. post test) where should the control mechanism appear in the loop? for counter-controlled loops • what are legal types and the scope of the loop control variable? • what happens to the variable when the loop terminates? • can the loop’s controlling variables (terminating value, step-size) be altered during execution? • are the loop controlling variables (terminating value, step-size) evaluated once (before executing the loop) or after each iteration? • what are legal step-sizes (for counter-controlled loops)

FORTRAN’s DO statement • DO label variable = initial, terminal [, step] – – example: Do 10 K = 1, 10 FORTRAN I – IV: a posttest loop, stepsize defaults to 1 label is a line number that indicates the last instruction in the loop body FORTRAN 77, 90 and 95: pretest loop • Integers (literals or variables) only for initial, terminal, step – these values are computed prior to loop execution so that, if a variable changes values in the loop, it does not affect the number of loop iterations DO 10 I = J, K * 10, L K=K+1 L=L+2 10 CONTINUE If J = 1, K = 5, L = 2, the loop would iterate 25 times in spite of K and L changing in the loop body initvalue = J terminalvalue = K * 10 stepvalue = L iterationcount = max(int((K*10 – J) / L), 1*) * - in FORTRAN I-IV, this is a post-test loop so it must iterate at least 1 time, in later FORTRANs, this would become 0

ALGOL For Loop • ALGOL 60 introduced an extremely flexible for loop as a reaction to FORTRAN’s primitive and restrictive Do – the programmer controls the number of iterations by • a counter controlled mechanism like FORTRAN’s DO but where the step size and terminating value could change during iterations • enumerating a list of values to iterate through • using a logical statement to control termination • or any combination thereof • Basic form: – for <var> : = <list>{, <list>} | <list> <expr> | <expr> while <boolean> | <expr> step <expr> until <expr> do <stmt> – examples: • • for count : =1, 2, 3, 4, 5, 6, 7 do list[count]: =0 for count: = 1 step 1 until 7 do list[count]: =0 for count: =1, count+1 while (count <=7) do list[count]: =0 for I : = 1, 4, 13, step 5 until 23, 3*I while I < 300, 8, -4 do … – the values for I iterate through: 1, 4, 13, 18, 23, 69, 207, 8, -4

Other Languages For Loops • COBOL: – Perform <expr> Times <statements> End-Perform – Perform Varying <var> From <expr> By <expr> Until <expr> <statements> End-Perform • PL/I: – DO <var> = <start> TO <stop> {BY <stepsize>}; <statements> END; – <start>, <stop> and <stepsize> can be int or float values – like FORTRAN though, the values are only evaluated once before the loop starts – can have multiple lists of <start> TO <stop> values • DO I = 1 TO 10, 20 to 30, 50 TO 100; … • Pascal: for <var> : = <init> (to | downto) <final> do – <var>, <init>, <final> are any ordinal type but are evaluated prior to the start of the loop, step size is fixed as 1 or -1 depending on whether you use to or downto • Ada: for <var> in [reverse] <range> loop. . . end loop – <range> is a subrange as in 1. . 10 (the values can be ints or enumerated types)

Continued • Common Lisp: (do ((<var> <init> <step>) (<endtest> {. <result>})) <statements>) – <init>, <step>, <endtest> and <result> can all be functions or atoms, <result> if specified is returned when the loop exits rather than the value returned by the last <statements> and <var>’s scope is only for the loop itself • C’s for loop – for (expr 1; expr 2; expr 3) statement • expr 1 is the initialization, expr 2 is the test, expr 3 is the step increment – each of these can be multiple terms separated by commas as in • for (c 1=0, c 2=1; c 1<=10 && c 2<=100; c 1++, c 2*=2) – notice a C for-loop does not need a loop body as actions can take place in the expr 3 component, or can omit one or more clauses, and can also be used like a logical loop • for(x=1; x<=n; x++, factorial *= x); • for(temp = head; temp != NULL; temp = temp->next) {…} – unlike the previous languages (except for Algol and Common Lisp), the increment and terminating conditions can change making these loops more writable but less readable

Iterator Loops • A variation of the counting loop is a loop that iterates once for each item in the list (or data structure) provided – Algol’s for loop has the capability of iterating over a list, but the list must be explicitly enumerated – Python: for <var> in <range>: • <range> will be a tuple or range(value [, value]) • note that Python’s for loop can also be a counting loop by using the range function as in for x in range(0, 10, 2) which iterates over 0, 2, 4, 6, 8, 10 – C# has a foreach statement which can iterate across array elements – Java 5. 0’s for loop has been enhanced to work on objects of type Iterable – Common Lisp has a dolist statement much like C#’s foreach

Logically Controlled Loops • For situations where the number of repetitions is not based on counting, we use logically controlled loops • Issues – pretest vs. posttest (test condition before entry or after execution? ) • pre-test can block entry to loop body • post-test must execute body at least once • in C, C++ and Java, the post-test loop has the same semantics as the pre-test loop: repeat while the condition is true • in Pascal, the semantics change: repeat until condition becomes false – is this type of statement separate from a special kind of counter-controlled loop? • C/C++/Java/C#, Pascal, Modula-2 have both pretest and posttest loops • Ada only has posttest • FORTRAN has no logically controlled loop (even FORTRAN 95)

Exiting Loops • Should exiting a loop only be permitted at the end when the test returns false, or can premature exiting (and returning) be permitted? – Ada has a conditional-less loop (infinite loop) • loop. . . end loop – to use this, it requires that a GOTO statement be used to break out of the loop, for instance in an if statement – C/C++/Java/C# and Modula-2 have unconditional exit statements • break, continue, exit – these are forms of GO TO statements • Java has break and exit but not continue – COBOL uses • Perform <paragraph> Thru <paragraph> – both paragraphs are executed but if an error arises in the first paragraph, control exits to the second paragraph • Multiple exits harm readability – exception throwing (and catching) are forms of pre-maturely exiting a block (including possibly inside a loop) • again, these are forms of GO TO statements

Problems with Unconditional Branching • Can make programs unreadable • Creates problems with maintenance – thus harming reliability, especially of very large programs • The GO TO statement is too primitive – “it is an invitation to make a mess of one’s program” • Most languages have some form of GOTO statement – Modula-2, Bliss, CLU, Euclid, Gypsy do not! – Java has a GOTO, but it hasn’t been implemented (yet) • Without a GOTO, the language must have other control mechanisms, usually in the form of loops and subprograms with their own ability to enter and exit – if you think about it, GOTO statements are required in assembly/machine code because they do not have the high-level language constructs, but the high level languages should offer these constructs and let the compiler do the work

GOTO Labels • Used in ALGOL 60, C, FORTRAN and Ada as locations for GOTO commands – in Ada, <<label>> – in FORTRAN and Pascal, unsigned integer constant (20, 100, etc. . . ) • in Pascal, labels must be declared as if they were variables (but cannot be modified or passed as a parameter) – in C, ALGOL 60, any legal identifier • Most languages restrict the use of unconditional branches with respect to which label can be reached – in Pascal, the scope of the label is the same as the scope of a variable and the target of the GOTO must be • within the control statement that includes the GOTO or • a statement in the same group that contains the GOTO or • in a statement in an enclosing subprogram scope