1. Data Types simple and compound abstraction and implementation

2. A brief history  simple types  arrays - compounds of same type  strings  records – compounds of different types  pointers and references  user defined types  abstract data types  objects

3. Simple types integer floating pointbinary-coded decimal character boolean user-defined types usually in hardware usually in software  not composed of other types  hardware or software implemented

4. Integer  2’s complement  unsigned  operations exact within range  range depends on size of virtual cell - typical size: 1, 2, 4, 8 bytes

5. Floating Point  based on scientific notation  representations and operations are approximate  range and precision depend on size of virtual cell (usually 4 or 8 bytes) 11152 bits

6. Binary Coded Decimal  ‘exact’ decimal arithmetic  decimal digits in 4 bit code  range and precision depend on size of virtual cell – 2 digits per byte 4 4 5905 18 78 defined decimal point

7. Character  ASCII – 128 character set – 1 byte  Unicode – 2 byte extension  usually coded as unsigned integer

8. Boolean  1 bit is sufficient but...  no bit-wise addressability in hardware  store in a byte – space inefficient  store 8 per byte – execution inefficient  c: 0=false, non-zero=true

9. User-defined types  implemented (like character and boolean usually are) as a coding of unsigned integer  enumerated type: (Pascal example) type suit = (club, diamond, heart, spade); var lead: suit; lead := heart; internally represented as { 0, 1, 2, 3 } operations:

10. User-defined types  implemented as a restricted range of integer  subrange type: (Ada example) subtype CENTURY20 is INTEGER range 1900..1999; BIRTHYEAR: CENTURY20; BIRTHYEAR := 1981;

11. User-defined types  Type compatibility issues: -can two enumerated types contain same constant? -can defined types be coerced with integer, with each other?

12. Memory management intro  The parser creates a symbol table of identifiers including variables: Some information, name plus more, is bound at this time and as the program is compiled by storage in symbol table: e.g. int x; --> xtype: int addr: offset name type address

13. Strings  First use: output formatting only  Quasi-primitive type in most languages (not just arrays of character) - operations: initialization, substring, catenation, comparison  The length problem: fixed or varying?  No standard string model

14. c char *s = “abc”; int len = strlen(s); array of char with terminal: extended syntax library of methods Strings - examples JAVA String s = “abc”+x; s = s.substring(0,2); fixed length array extended syntax class with 70 methods a b c 0

15. Strings - representations  fixed length and content (static)  fixed length and varying content (FORTRAN)  varying length and content by reallocation (java String)  varying length and content by extension (java StringBuffer)  Varying length and content(c) Static str Length Address Dynamic str MaxLength CurrLength Address char* Address In symbol table

16. Compound (1) Arrays  collection of elements of one type  access to individual elements is computed at execution time by position, O(1), or O(dim)

17. Arrays – design decisions  indexing: dimensions – limit? recursive? types – int, other, user defined? first index: 0, 1, variable range checking – no(c), yes(java) lexemes – ‘subscripts’ = (),[]?

18. Arrays – design decisions  binding times type, index type index range(ie array size), space static fixed stack-dynamic stack-dynamic heap-dynamic initial values of elements at storage allocation? e.g. int[] x = {1,2,3};

19. Arrays – operations  on elements – based on type  on entire array as variables - - vector and matrix operations e.g.,APL - sub array (~ substring)  subarray dimensions(slices)

20. Arrays – storage element type, size index type index lower bound index upper bound address lower bound upper bound

21. Arrays – element access element type, size index type index lower bound index upper bound address lower bound i address of a[i] = address + (i-lower bound)*size

22. Arrays - multidimensional  contiguous or not  row major, column major order  computed location of element

23. Jagged arrays  Implemented as arrays of arrays, 4 index type index lower bound index upper bound address, 3 index type index lower bound index upper bound address, 7 index type index lower bound index upper bound address, 4 index type index lower bound index upper bound address, 5 index type index lower bound index upper bound address

24. (2) Associative Arrays - maps  values accessed by keys,not indices  no order of elements  automatic growth of capacity  operations: add/set, get, remove  fast search for individual data  slower for batch processing than array  Java classes; Perl data structure

25. Associative Arrays - implementation  hash tables based on key value  most operations ‘near O(1)’  expanding capacity may be O(n) For a java class that combines features of array and associative array, see LinkedHashMap

26. (3) Records  multiple elements of any type  elements accessed by field name  design issues: -hierarchical definition (records within records) -syntax of naming -scopes for elliptical (incomplete) reference to fields

27. Records - implementation a element type, size index type index lower bound index upper bound address lower bound upper bound dept array [1..4] of char 0 (offset) code address C OSC3127 dept course integer 4 type course = record dept : array[1..4] of char; code : integer; end

28. (4) Pascal variant records (unions) type coord = (polar, cart); point = record case rep : coord of polar: ( radians : boolean; radius : real; angle : real); cart: ( x : real; y : real); end; Note: varying space requirements discriminant field is optional (rep) type checking loopholes: Ada has similar variant record but closed these loopholes

29. Other unions  Fortran EQUIVALENCE  c union  not inside records  no type checking * unions do not cause type coercion - data is reinterpreted Sebesta’s c example union flextype { int intE1; float floatE1; } union flexType ell; float x; ell.intE1 = 27; x = ell.floatE1; Sebesta’s c example union flextype { int intE1; float floatE1; } union flexType ell; float x; ell.intE1 = 27; x = ell.floatE1;

30. (5) Sets (Pascal)  defined on one (discrete) base type  implementation imposes maximum size ( set of integer;- not possible) type day = (M, Tu, W, Th, F, Sa, Su); dayset = set of day; var work, wknd : dayset; today : day; today = F; work = [M, Tu, W, Th, F]; wknd = [Sa, Su, F]; if (today in work and wknd)... 1101110 0011001 0001000

31. (6) Pointers and references  references are dereferenced pointers (whatever that means)  primary purpose: dynamic memory access  secondary purpose: indirect addressing as in machine instructions

32. Pointers (and references)  data type that stores an address in the format of the machine (usually 4 bytes) or a “null”  a pointer must be dereferenced to get the data at the address it contains  a reference is a pointer data type that is automatically dereferenced

33. Dereferencing example In c++: double x,y; Point p(0.0,0.0); Point *pref; pref = &p; x = p.X; y = (*pref).Y; In Java: Point2D.Double p; p = new Point2D.Double(0.0,0.0); double xCoord = p.x; Dereferencing and field access combined DereferencingField access

34. Pointers hold addresses  Indirect addressing In c: pointer to statically allocated memory int a,b; int *iptr, *jptr; a = 100; iptr = &a; jptr = iptr; b = *jptr; int x, y, arr[4]; int *iptr; iptr = arr; arr[2] = 33; x = iptr[2]; y = *(iptr + 2);  Security loophole…

35. Pointer arithmetic  Arithmetic operations on addresses int x; int *iptr; iptr = &x; for (;;){ > iptr++; } Scan through memory starting at x

36. Basic dynamic memory management model:  Heap manager keeps list of available memory cells  “Allocate” operation transfers cell from list in heap to program  “Deallocate” transfers cell from program back to list in heap  Tradeoffs of fixed or variable sized cells

37. Problems with pointers and dynamic memory:1  Dangling reference: pointer points to de-allocated memory Point *q; Point *p = new Point(0,0); q = p; delete p; // q is dangling - reference to q should cause // an error - ‘tombstones’ will do error check

38. Problems with pointers and dynamic memory: 2  Memory leakage: memory cell with no reference to it Point *p = new Point(0,0); p = new Point(3,4); // memory containing Point(0,0) object // is inaccessible - counting references will help

39. Cause of reference problems  Multiple references to a memory cell  Deallocation of memory cells  Where is responsibility? -automatic deallocation (garbage collection) OR -user responsibility (explicit ‘delete’)

40. User management of memory  Dangling references can be detected as errors but not prevented - tombstones - lock and key  Memory leakage is a continuing problem int *p =*q = 6; p = null; int *p =*q = 6; p = null; p 6 q p 6 q

41. Garbage Collection 1.Reference counting:ongoing “eager” -memory cells returned to heap as soon as all references removed. 2.Garbage collection:occasional “lazy” -let unreferenced memory cells ‘leak’ till heap is nearly empty then collect them

42. Reference counting: 2 p = null; 1 0 q = null; Reference count in cell Count 0 -> return cell to heap Classic problem: circular linked lists int *p = *q = 6;

43. Garbage Collection: (mark-sweep) 1. All cells in memory marked inaccessible(f) 2. Follow all references in program and mark cells accessible(t); f t t ‘Accessible’ marker in cell 3. Return inaccessible cells to heap f t t Classic problem: effect on program performance

44. A sloppy java example  from Main (Data Structures) public class ObjectStack { private Object[] data; private int manyItems;.... public Object pop() { if (manyItems==0) throw new EmptyStackException(); return data[--manyItems]; //leaves reference in data }

45. Managing heap of variable-sized cells  Necessary for objects with different space requirements  Problem: tracking cell size  Problem: heap defragmentation - keep blocks list in size order? - keep blocks list in sequence order?