Notes

Outline

Object-Oriented and Classical Software Engineering   Fifth Edition, WCB/McGraw-Hill, 2002 Stephen R. Schach srs@vuse.vanderbilt.edu

CHAPTER 11

Overview The specification document Informal specifications Structured systems analysis Other semiformal techniques Entity-relationship modeling Finite state machines Petri nets Other formal techniques

Overview (contd) Comparison of specification techniques Testing during the specification phase CASE tools for the specification phase Metrics for the specification phase Air Gourmet Case Study: structured systems analysis Air Gourmet Case Study: software project management plan Challenges of the specifications phase

Specification Phase Specification document must be Informal enough for client Formal enough for developers Free of omissions, contradictions, ambiguities

Specification Document Constraints Cost Time Parallel running Portability Reliability Rapid response time

Specification Document (contd) Acceptance criteria Vital to spell out series of tests Product passes tests, deemed to satisfy specifications Some are restatements of constraints

Solution Strategy General approach to building the product Find strategies without worrying about constraints Modify strategies in the light of constraints, if necessary  Keep a written record of all discarded strategies, and why they were discarded To protect the specification team To prevent unwise new “solutions” during the maintenance phase

Informal Specifications Example “If sales for current month are below target sales, then report is to be printed, unless difference between target sales and actual sales is less than half of difference between target sales and actual sales in previous month, or if difference between target sales and actual sales for the current month is under 5%”

Meaning of Specification Sales target for January was $100,000, actual sales were only $64,000 (36% below target) Print report Sales target for February was $120,000, actual sales were only $100,000 (16.7% below target) Percentage difference for February (16.7%) less than half of previous month’s percentage difference (36%), do not print report Sales target for March was $100,000, actual sales were $98,000 (2% below target) Percentage difference < 5%, do not print

But Specifications Do Not Say This “[D]ifference between target sales and actual sales” There is no mention of percentage difference Difference in January was $36,000, difference in February was $20,000 Not less than half of $36,000, so report is printed “[D]ifference … [of] 5%” Again, no mention of percentage Ambiguity—should the last clause read “percentage difference … [of] 5%” or “difference … [of] $5,000” or something else entirely? Style is poor

Informal Specifications (contd) Claim This cannot arise with professional specifications writers Refutation Text Processing case study

Episode 1 1969 — Naur Paper Given a text consisting of words separated by blank or by nl (new line) characters, convert it to line-by-line form in accordance with following rules: (1) line breaks must be made only where given text has blank or nl ; (2) each line is filled as far as possible, as long as (3) no line will contain more than maxpos characters Naur constructed a procedure (25 lines of Algol 60), and informally proved its correctness

Episode 2 1970 — Reviewer in Computing Reviews First word of first line is preceded by a blank unless the first word is exactly maxpos characters long

Episode 3 1971 — London found 3 more faults Including: procedure does not terminate unless a word longer than maxpos characters is encountered

Episode 4 1975 — Goodenough and Gerhart found 3 further faults Including—last word will not be output unless it is followed by blank or nl Goodenough and Gerhart then produced new set of specifications, about four times longer than Naur’s

Case Study (contd) 1985 — Meyer detected 12 faults in Goodenough and Gerhart’s specifications Goodenough and Gerhart’s specifications Were constructed with the greatest of care Were constructed to correct Naur’s specifications Went through two versions, carefully refereed Were written by experts in specifications With as much time as they needed For a product about 30 lines long What chance do we have of writing fault-free specifications for a real product?

Episode 5 1989 — Schach found fault in Meyer’s specifications Item (2) of Naur’s original requirement (“each line is filled as far as possible”) is not satisfied

Informal Specifications Conclusion Natural language is not a good way to specify product Fact Many organizations still use natural language, especially for commercial products Reasons Uninformed management Undertrained computer professionals Management gives in to client pressure Management is unwilling to invest in training

Structured Systems Analysis Three popular graphical specification methods of ’70s DeMarco Gane and Sarsen Yourdon All equivalent All equally good Many U.S. corporations use them for commercial products Gane and Sarsen used for object-oriented design

Structured Systems Analysis Case Study Sally’s Software Store buys software from various suppliers and sells it to the public.  Popular software packages are kept in stock, but the rest must be ordered as required.  Institutions and corporations are given credit facilities, as are some members of the public.  Sally’s Software Store is doing well, with a monthly turnover of 300 packages at an average retail cost of $250 each.  Despite her business success, Sally has been advised to computerize.  Should she? Better question What sections? Still better How? Batch, or online?  In-house or out-service?

Case Study (contd) Fundamental issue What is Sally’s objective in computerizing her business? Because she sells software? She needs an in-house system with sound and light effects Because she uses her business to launder “hot” money? She needs a product that keeps five different sets of books, and has no audit trail Assume: Computerization “in order to make more money” Cost/benefit analysis for each section of business

Case Study (contd) The danger of many standard approaches First produce the solution, then find out what the problem is! Gane and Sarsen’s method Nine-step method Stepwise refinement is used in many steps

Case Study (contd) Data flow diagram (DFD) shows logical data flow “what happens, not how it happens”

Step 1.  Draw the DFD First refinement Infinite number of possible interpretations

Step 1 (contd) Second refinement pending orders scanned daily

Step 1 (contd) Portion of third refinement

Step 1 (contd) Final DFD Larger, But easily understood by client Larger DFDs Hierarchy Box becomes DFD at lower level Frequent problem Process P at level L, expanded at level L+1 Correct place for sources and destinations of data for process P is level L+1 Clients cannot understand DFD—sources and destinations of data for P are “missing” Solution Draw “correct” DFD, modify by moving sources and destinations of data one or more levels up

Step 2. Decide What Parts to Computerize Depends on how much client is prepared to spend Large volumes, tight controls Batch Small volumes, in-house microcomputer Online Cost/benefit analysis

Step 3. Refine Data Flows Data items for each data flow Refine each flow stepwise Refine further Need a data dictionary

Step 3. Refine Data Flows (contd) Sample data dictionary entries

Step 4.   Refine Logic of Processes Have process give educational discount Sally must explain discount for educational institutions 10% on up to 4 packages, 15% on 5 or more Translate into decision tree

Step 4 (contd) Advantage of decision tree Missing items are quickly apparent Can also use decision tables CASE tools for automatic translation

Step 5.   Refine Data Stores Define exact contents and representation (format) COBOL: specify to pic level Ada: specify digits or delta Specify where immediate access is required Data immediate access diagram (DIAD)

Step 6.  Define Physical Resources For each file, specify File name Organization (sequential, indexed, etc.) Storage medium Blocking factor Records (to field level)

Step 7.  Determine Input/Output Specs Specify input forms, input screens, printed output

Step 8.  Perform Sizing Numerical data for Step 9 to determine hardware requirements Volume of input (daily or hourly) Size, frequency, deadline of each printed report Size, number of records passing between CPU and mass storage Size of each file

Step 9.  Hardware Requirements DASD requirements Mass storage for back-up Input needs Output devices Is existing hardware adequate? If not, recommend buy/lease

However Response times cannot be determined Number of I/O channels can only be guessed CPU size and timing can only be guessed Nevertheless, no other method provides these data for arbitrary products The method of Gane and Sarsen/De Marco/Yourdon has resulted in major improvements in the software industry

Entity-Relationship Diagrams Semi-formal technique Widely used for specifying databases Used for object-oriented analysis

Entity-Relationship Diagrams (contd) Many-to-many relationship More complex entity-relationship diagrams

Formality versus Informality Informal method English (or other natural language) Semiformal methods Gane & Sarsen/DeMarco/Yourdon Entity-Relationship Diagrams Jackson/Orr/Warnier, SADT, PSL/PSA, SREM, etc. Formal methods Finite State Machines Petri Nets Z ANNA, VDM, CSP, etc.

Finite State Machines Case study A safe has a combination lock that can be in one of three positions, labeled 1, 2, and 3.  The dial can be turned left or right (L or R).  Thus there are six possible dial movements, namely 1L, 1R, 2L, 2R, 3L, and 3R.  The combination to the safe is 1L, 3R, 2L; any other dial movement will cause the alarm to go off

Finite State Machines (contd) Transition table

Extended Finite State Machines Extend FSM with global predicates Transition rules have form state and event and predicate  Þ  new state

Elevator Problem A product is to be installed to control n elevators in a building with m floors.  The problem concerns the logic required to move elevators between floors according to the following constraints: 1. Each elevator has a set of m buttons, one for each floor.  These illuminate when pressed and cause elevator to visit corresponding floor.  Illumination is canceled when corresponding floor is visited by elevator 2. Each floor, except the first and the top floor, has 2 buttons, one to request an up-elevator, one to request a down-elevator.  These buttons illuminate when pressed.  The illumination is canceled when an elevator visits the floor, then moves in the desired direction 3. If an elevator has no requests, it remains at its current floor with its doors closed

Elevator Problem: FSM Two sets of buttons Elevator buttons—in each elevator, one for each floor Floor buttons—two on each floor,  one for up-elevator, one for down-elevator EB(e, f):  Elevator Button in elevator e pressed to request floor f

Elevator Buttons (contd) Two states EBON(e, f): Elevator Button (e,f) ON EBOFF(e,f): Elevator Button (e,f) OFF If button is on and elevator arrives at floor f, then light turned off If light is off and button is pressed, then light comes on

Elevator Buttons (contd) Two events EBP(e,f): Elevator Button (e,f) Pressed EAF(e,f): Elevator e Arrives at Floor f Global predicate V(e,f): Elevator e is Visiting (stopped at) floor f Transition Rules EBOFF(e,f) and EBP(e,f) and not V(e,f) Þ EBON(e,f) EBON(e,f) and EAF(e,f) Þ EBOFF(e,f)

Floor Buttons Floor buttons FB(d, f): Floor Button on floor f that requests elevator traveling in direction d States FBON(d, f): Floor Button (d, f) ON FBOFF(d, f): Floor Button (d, f) OFF If floor button is on and an elevator arrives at floor f, traveling in correct direction d, then light is turned off If light is off and a button is pressed, then light comes on

Floor Buttons (contd) Events FBP(d, f): Floor Button (d, f) Pressed EAF(1..n, f): Elevator 1 or … or n Arrives at Floor f Predicate S(d, e, f): elevator e is visiting floor f Direction of motion is up (d = U), down (d = D), or no requests are pending (d = N) Transition rules FBOFF(d, f) and FBP(d, f) and not S(d, 1..n, f) Þ FBON(d, f) FBON(d, f) and EAF(1..n, f) and S(d, 1..n, f) Þ FBOFF(d, f), d = U or D

Elevator Problem: FSM (contd) State of elevator consists of component substates, including: Elevator slowing Elevator stopping Door opening Door open with timer running Door closing after a timeout

Elevator Problem: FSM (contd) Assume elevator controller moves elevator through substates Three elevator states M(d, e, f): Moving in direction d (floor f is next) S(d, e, f): Stopped (d-bound) at floor f W(e,f): Waiting at floor f (door closed) For simplicity, three stopped states S(U, e, f), S(N, e, f), and S(D, e, f) are grouped into one larger state

Elevator Problem: FSM (contd)

Elevator Problem: FSM (contd) Events DC(e,f): Door Closed for elevator e, floor f ST(e,f): Sensor Triggered as elevator e nears floor f RL: Request Logged (button pressed) Transition Rules If elevator e is in state S(d, e, f) (stopped, d-bound, at floor f), and doors close, then elevator e will move up, down, or go into wait state DC(e,f) and S(U, e, f) Þ M(U, e, f+1) DC(e,f) and S(D, e, f) Þ M(D, e, f-1) DC(e,f) and S(N, e, f) Þ W(e,f)

Power of FSM to Specify Complex Systems No need for complex preconditions and postconditions Specifications take the simple form current state and event and predicate Þ next state

Power of FSM to Specify Complex Systems Using an FSM, a specification is Easy to write down Easy to validate Easy to convert into design Easy to generate code automatically More precise than graphical methods Almost as easy to understand Easy to maintain However Timing considerations are not handled

Who Is Using FSMs? Commercial products Menu driven Various states/screens Automatic code generation a major plus System software Operating system Word processors Spreadsheets Real-time systems Statecharts are a real-time extension of FSMs CASE tool: Rhapsody

Petri Nets A major difficulty with specifying real-time systems is timing Synchronization problems Race conditions Deadlock Often a consequence of poor specifications

Petri Nets (contd) Petri nets Powerful technique for specifying systems with potential timing problems A Petri net consists of four parts: Set of places P Set of transitions T Input function I Output function O

Petri Nets (contd) Set of places P is {p1, p2, p3, p4} Set of transitions T is {t1, t2} Input functions: I(t1) = {p2, p4} I(t2) = {p2} Output functions: O(t1) = {p1} O(t2) = {p3, p3}

Petri Nets (contd) More formally, a Petri net is a 4-tuple C = (P, T, I, O) P = {p1, p2,…,pn} is a finite set of places, n ≥ 0 T = {t1, t2,…,tm} is a finite set of transitions, m ≥ 0, with P and T disjoint I : T ® P∞ is input function, mapping from transitions to bags of places O : T ® P∞ is output function, mapping from transitions to bags of places (A bag is a generalization of sets which allows for multiple instances of element in bag, as in example above) Marking of a Petri net is an assignment of tokens to that Petri net

Petri Nets (contd) Four tokens, one in p1, two in p2, none in p3, and one in p4 Represented by vector (1,2,0,1) Transition is enabled if each of its input places has as many tokens in it as there arcs from the place to that transition

Petri Nets (contd) Transition t1 is enabled (ready to fire) If t1 fires, one token is removed from p2 and one from p4, and one new token is placed in p1 Important: Number of tokens is not conserved Transition t2 is also enabled

Petri Nets (contd) Petri nets are indeterminate Suppose t1 fires Resulting marking is (2,1,0,0)

Petri Nets (contd) Now only t2 is enabled It fires Marking is now (2,0,2,0)

Petri Nets (contd) More formally, a marking M of a  Petri net           C = (P, T, I, O) is a function from the set of places P to the non-negative integers N M : P ® N A marked Petri net is then 5-tuple (P, T, I, O, M )

Petri Nets (contd) Inhibitor arcs Inhibitor arc is marked by small circle, not arrowhead Transition t1 is enabled In general, transition is enabled if at least one token on each (normal) input arc, and no tokens on any  inhibitor input arcs

Elevator Problem: Petri Net Product is to be installed to control n elevators in a building with m floors Each floor represented by place Ff, 1  £  f £ m Elevator represented by token Token in Ff denotes that an elevator is at floor Ff

Elevator Problem: Petri Net (contd) First constraint 1. Each elevator has a set of m buttons, one for each floor.  These illuminate when pressed and cause the elevator to visit the corresponding floor.  The illumination is canceled when the corresponding floor is visited by an elevator Elevator button for floor f is represented by place EBf, 1  £ f £ m Token in EBf denotes that the elevator button for floor f is illuminated

Elevator Problem: Petri Net (contd) A button must be illuminated the first time button is pressed and subsequent button presses must be ignored If button EBf is not illuminated, no token in place and transition EBf pressed is enabled Transition fires, new token is placed in EBf Now, no matter how many times button is pressed, transition EBf pressed cannot be enabled

Elevator Problem: Petri Net (contd) When elevator reaches floor g, token is in place Fg, transition Elevator in action is enabled, and then fires Tokens in EBf and Fg removed This turns off light in button EBf New token appears in Ff This brings elevator from floor g to floor f

Elevator Problem: Petri Net (contd) Motion from floor g to floor f cannot take place instantaneously Timed Petri nets

Elevator Problem: Petri Net (contd) Second constraint 2. Each floor, except the first and the top floor, has 2 buttons, one to request an up-elevator, one to request a down-elevator.  These buttons illuminate when pressed.  The illumination is canceled when the elevator visits the floor, and then moves in desired direction Floor buttons represented by places FBuf and FBdf

Elevator Problem: Petri Net (contd)

Elevator Problem: Petri Net (contd) The situation when an elevator reaches floor f from floor g with one or both buttons illuminated If both buttons are illuminated, only one is turned off (A more complex model is needed to ensure that the correct light is turned off)

Elevator Problem: Petri Net (contd) Third constraint C3. If an elevator has no requests, it remains at its current floor with its doors closed If no requests, no Elevator in action transition is enabled

Petri Nets (contd) Petri nets can also be used for design Petri nets possess the expressive power necessary for specifying timing aspects of real-time systems

Z (“zed”) Formal specification language High squiggle factor Z specification consists of four sections: 1. Given sets, data types, and constants 2. State definition 3. Initial state 4. Operations

Elevator Problem: Z 1.  Given Sets Sets that need not be defined in detail Names appear in brackets Here we need the set of all buttons Specification begins        [Button]

Elevator Problem: Z (contd) 2.  State Definition Z specification consists of a number of schemata Schema consists of group of variable declarations, plus List of predicates that constrain values of variables

Elevator Problem: Z (contd) Four subsets of Button The floor buttons The elevator buttons buttons (the set of all buttons in the elevator problem) pushed (the set of buttons that have been pushed)

Elevator Problem: Z (contd) Schema Button_State

Elevator Problem: Z (contd) 3.  Initial State State when the system is first turned on        Button_Init º [Button_State' | pushed' = Æ] (In the above equation, the º should be a = with a ^ on top.  Unfortunately, this is hard to type in PowerPoint!)

Elevator Problem: Z (contd) 4.  Operations Button pushed for first time is turned on, and added to set pushed Without third precondition, results would be unspecified

Elevator Problem: Z (contd) If elevator arrives at a floor, the corresponding button(s) must be turned off The solution does not distinguish between up and down floor buttons

Analysis of Z Most widely used formal specification language CICS (part) Oscilloscope CASE tool Large-scale projects (esp. Europe)

Analysis of Z (contd) Difficulties Symbols Mathematics Reasons for great success Easy to find faults in Z specification Specifier must be extremely precise Can prove correctness Only high-school math needed to read Z Decreases development time “Translation” clearer than informal specification

Other Formal Methods Anna Ada Gist, Refine Knowledge-based VDM Denotational semantics CSP Sequence of events Executable specifications High squiggle factor

Comparison of Specification Techniques We must always choose the appropriate specification method Formal methods Powerful Difficult to learn and use Informal methods Little power Easy to learn and use Trade-off   Ease of use versus power

Comparison of Specification Techniques (contd)

Newer Methods Many are untested in practice Risks Training costs Adjustment from classroom to actual project CASE tools may not work properly However, possible gains may be huge

Which Specification Method to Use? Depends on the Project Development team Management team Myriad other factors It is unwise to ignore the latest developments

Testing during the Specification Phase Specification inspection Checklist Doolan [1992] 2 million lines of FORTRAN 1 hour of inspecting saved 30 hours of execution-based testing

CASE Tools for the Specification Phase Graphical tool Data dictionary Integrate them Specification method without CASE tools fails SREM

CASE Tools for the Specification Phase Typical tools Analyst/Designer Software through Pictures System Architect

Metrics for the Specification Phase Five fundamental metrics Quality Fault statistics Number, type of each fault Rate of detection Metrics for “predicting” size of target product Total number of  items in data dictionary Number of items of each type Processes vs. modules

Air Gourmet Case Study: Structured Sys. Anal. Data flow diagram reflects centrality of SPECIAL MEAL DATA See Appendix E for remainder of structured systems analysis

Air Gourmet Case Study: SPMP The Software Project Management Plan is given in Appendix F

Challenges of the Specification Phase A specification document must be Informal enough for the client; and Formal enough for the development team The specification phase (“what”) should not cross the boundary into the design phase (“how”) Do not try to assign modules to process boxes of DFDs until the design phase