IF

IF

Execute one of several alternative blocks of statements

Syntax

       IF <lCondition1>
          <statements>...
       [ELSEIF <lCondition2>]
          <statements>...
       [ELSE]
          <statements>...
       END[IF]

Arguments

<lCondition> is a logical control expression. If it evaluates to true (.T.), all following statements are executed until an ELSEIF, ELSE, or ENDIF is encountered.

ELSEIF <lCondition> identifies statements to execute if the associated condition evaluates to true (.T.) and all preceding IF or ELSEIF conditions evaluate to false (.F.). Any number of ELSEIF statements can be specified within the same IF…ENDIF control structure.

ELSE identifies statements to execute if the IF and all preceding ELSEIF conditions evaluate to false (.F.).

Description

The IF control structure works by branching execution to statements following the first true (.T.) evaluation of the IF or any ELSEIF condition. Execution then continues until the next ELSEIF, ELSE, or ENDIF is encountered whereupon execution branches to the first statement following the ENDIF.

If no condition evaluates to true (.T.), control passes to the first statement following the ELSE statement. If an ELSE statement is not specified, control branches to the first statement following the ENDIF statement.

IF…ENDIF structures may be nested within IF…ENDIF structures and other control structure commands. These structures, however, must be nested properly. The IF…ELSEIF…ENDIF form of the IF construct is identical to DO CASE…ENDCASE. There is no specific advantage of one syntax over the other.

The IF construct is also similar to the IF() function which can be used within expressions.

Examples

       .  This example evaluates a number of conditions using an
       IF...ELSEIF...ENDIF construct:
       LOCAL nNumber := 0
       //
       IF nNumber < 50
          ? "Less than 50"
       ELSEIF nNumber = 50
          ? "Is equal to 50"
       ELSE
          ? "Greater than 50"
       ENDIF

Seealso

BEGIN SEQUENCE, DO CASE, DO WHILE, FOR, IF()

FOR…NEXT

FOR…NEXT

Execute a block of statements a specified number of times

Syntax

       FOR <idCounter> := <nStart> TO <nEnd> [STEP <nIncrement>]
           <statements>...
           [EXIT]
           <statements>...
           [LOOP]
       NEXT

Arguments

<idCounter> is the name of the loop control or counter variable. If the specified <idCounter> is not visible or does not exist, a private variable is created.

<nStart> is the initial value assigned to <idCounter>. If <nIncrement> is negative, <nStart> must be less than <nEnd>.

TO <nEnd> defines the final value of <idCounter>. If <nIncrement> is negative, <nStart> must be greater than <nEnd>; otherwise, <nStart> must be less than <nEnd>.

STEP <nIncrement> defines the amount <idCounter> is changed for each iteration of the loop. <nIncrement> can be either positive or negative. If the STEP clause is not specified, <idCounter> is incremented by one for each iteration of the loop.

EXIT unconditionally branches control from within a FOR…NEXT construct to the statement immediately following the nearest NEXT statement.

LOOP branches control to the most recently executed FOR or DO WHILE statement.

Description

FOR…NEXT is a control structure that executes a block of statements a specified number of times. The control structure loops from the initial value of <idCounter> to the boundary specified by <nEnd>, moving through the range of values of the control variable for an increment specified by <nIncrement>. All expressions in the FOR statement are reevaluated for each iteration of the loop. The <nStart> and <nEnd> expressions, therefore, can be changed as the control structure operates.

A FOR loop operates until <idCounter> is greater than <nEnd> or an EXIT statement is encountered. Control then branches to the statement following the corresponding NEXT statement. If a LOOP statement is encountered, control branches back to the current FOR statement.

If <nIncrement> is a negative value, <idCounter> is decremented rather than incremented. The FOR loop, however, continues until <idCounter> is less than <nEnd>. This means that <nEnd> must be less than <nStart> when the FOR loop begins.

FOR loops are useful for traversing arrays where <idCounter> is used as the array subscript. See the example below.

FOR…NEXT constructs may be nested within any other control structures to any depth. The only requirement is that each control structure is properly nested.

Examples

       .  This example traverses an array in ascending order:
       nLenArray := LEN(aArray)
       FOR i := 1 TO nLenArray
           <statements>...
       NEXT
       .  To traverse an array in descending order:
       nLenArray := LEN(aArray)
       FOR i := nLenArray TO 1 STEP -1
          <statements>...
       NEXT

Seealso

AEVAL(), BEGIN SEQUENCE, DO CASE, DO WHILE, IF, IF()

C5_IF ( Function )

 IF()
 Return the result of an expression based on a condition
------------------------------------------------------------------------------
 Syntax

     IF(<lCondition>, <expTrue>, <expFalse>) --> Value

 Arguments

     <lCondition> is a logical expression to be evaluated.

     <expTrue> is the value, a condition-expression, of any data type,
     returned if <lCondition> is true (.T.).

     <expFalse> is the value, of any date type, returned if <lCondition>
     is false (.F.).  This argument need not be the same data type as
     <expTrue>.

 Returns

     IF() returns the evaluation of <expTrue> if <lCondition> evaluates to
     true (.T.) and <expFalse> if it evaluates to false (.F.).  The value
     returned is the data type of the valid condition-expression.

 Description

     IF() is a logical conversion function.  It is one of the most powerful
     and versatile functions in Clipper.  It provides a mechanism to
     evaluate a condition within an expression.  With this ability you can
     convert a logical expression to another data type.

 Examples

     .  This example converts a logical data value to a numeric data
        value:

        lPaid = .T.
        ? IF(lPaid, 1, 0)               // Result: 1

     .  In this example a logical field is formatted depending on
        whether the Customer is past due or not:

        @ ROW() + 1, 25 SAY IF(lPaid, SPACE(10), "Go get 'em")

     .  If you are printing forms, you can print an indicating symbol
        in different columns depending on the value of a logical field:

        @ ROW(), IF(InHospital, 10, 12) SAY "X"

     .  You can also use IF() to force the LABEL FORM to print blank
        lines.  Enter the following expression when you create the label with
        RL.EXE:

        IF(EMPTY(Company), CHR(255), Company)

 Files   Library is CLIPPER.LIB.

Note : IF() is synonymous (shorthand)  of IIF().

See Also: DO CASE IF() IIF()



Clipper and Networking

Clipper and Networking

Hash Basics

Definition:

In general, a Hash Table, or Hash Array, or Associative array, or shortly Hash is an array- like data structure, to store some data with an associated key for each; so, ‘atom’ of a hash is a pair of a ‘key’ with a ‘value’. A hash system needs to perform at least three operations:

–      add a new pair,

–      access to value via key

–      the search and delete operations on a key pair

In Harbour, a hash is simply a special array, or more precisely a “keyed” array with special syntax with a set of functions.

Building:

The “=>” operator can be used to indicate literally the relation between <key> <value> pair: <key> => <value>

 We can define and initialize a hash by this “literal” way :

 hDigits_1 := { 1 => 1, 2  => 2, 3  => 3, 4  => 4 }

 or by a special function call:

 hDigits_1 := HB_HASH( 1, 1, 2, 2, 3, 3, 4, 4 )

 Using “add” method may be another way :

hDigits_1 := { => } // Build an empty hash
hDigits_1[ 1] := 1

hDigits_1[ 2] := 2

hDigits_1[ 3] := 3

hDigits_1[ 4] := 4

In this method while evaluating each of above assignments, if given key exits in hash, will be replaced its value; else add a new pair to the hash.

In addition, data can be added to a hash by extended “+=” operator:

   hCountries := { 'Argentina' => "Buenos Aires" }
   hCountries += { 'Brasil'    => "Brasilia" }
   hCountries += { 'Chile'     => "Santiago" }
   hCountries += { 'Mexico'    => "Mexico City" }

Hashs may add ( concatenate ) each other by extended “+” sign :

   hFruits := { "fruits" => { "apple", "chery", "apricot" } }
   hDays   := { "days"   => { "sunday", "monday" } } 
   hDoris := hFruits + hDays

Note:  This “+” and “+=” operators depends xHB lib and needs to xHB lib and xHB.ch.

Typing :

<key> part of a hash may be any legal scalar type : C, D, L, N; and <value> part may be in addition scalar types, any complex type ( array or hash ).

Correction : This definition is wrong ! The correct is :

<key> entry key; can be of type: number, date, datetime, string, pointer.

Corrected at : 2015.12.08; thanks to Marek.

hDigits_2 := {  1  => “One”,  2  => “Two”,  3  => “Three”,  4  => “Four” }

hDigits_3 := { "1" => "One", "2" => "Two", "3" => "Three", "4" => "Four" }
hDigits_4 := { "1" => "One",  2  => "Two",  3  => "Three", "4" => "Four" }
hDigits_5 := {  1  => "One",  1  => "Two",  3  => "Three",  4  => "Four"

All of these examples are legal. As a result, a pair record of a hash may be:

–      Numeric key, numeric value ( hDigits_1 )

–      Numeric key, character value ( hDigits_2 )

–      Character key, character value ( hDigits_3 )

–      Mixed type key ( hDigits_4 )

Duplicate keys (as seen in hDigits_5) is permitted to assign, but not give a result such as double keyed values: LEN( hDigits_5 ) is 3, not 4; because first pair replaced by second due to has same key.

Consider a table-like data for customers records with two character fields: Customer ID and customer name:

Cust_ID Cust_Name
CC001 Pierce Firth
CC002 Stellan Taylor
CC003 Chris Cherry
CC004 Amanda Baranski

We can build a hash with this data :

  hCustomers := { "CC001" => "Pierce Firth",;
 "CC002" => "Stellan Taylor",;
 "CC003" => "Chris Cherry",;
 "CC004" => "Amanda Baranski" }

and list it:

   ?
   ? "Listing a hash :"
   ?
   h1Record := NIL
   FOR EACH h1Record IN hCustomers
      ? cLMarj, h1Record:__ENUMKEY(), h1Record:__ENUMVALUE()
   NEXT

 Accessing a specific record is easy :

 hCustomers[ "CC003" ] // Chris Cherry
*-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.
/*
Hash Basics

*/
#include "xhb.ch"
#define NTrim( n ) LTRIM( STR( n ) )
PROCEDURE Main()
 SET DATE GERM
 SET CENT ON
 SET COLO TO "W/B"

 cLMarj := SPACE( 3 )

 CLS

 hDigits_1 := { => } // Build an empty hash

 hDigits_1[ 1 ] := 1
 hDigits_1[ 2 ] := 2
 hDigits_1[ 3 ] := 3
 hDigits_1[ 4 ] := 4

 ListHash( hDigits_1, "Digits_1" )

 hDigits_2 := HB_HASH( 1, 1, 2, 2, 3, 3, 4, 4 )

 ListHash( hDigits_2, "Digits_2" )

 hDigits_3 := { 1 => 1,;
 2 => 2,;
 3 => 3,;
 4 => 4 }
 ListHash( hDigits_3, "Digits_3" )

 hDigits_4 := { 1 => "One",;
 2 => "Two",;
 3 => "Three",;
 4 => "Four" }
ListHash( hDigits_4, "Digits_4" )

 hDigits_5 := { "1" => "One",;
 "2" => "Two",;
 "3" => "Three",;
 "4" => "Four" }
 ListHash( hDigits_5, "Digits_5" )

 hDigits_6 := { "1" => "One",;
 2 => "Two",;
 3 => "Three",;
 "4" => "Four" }
 ListHash( hDigits_6, "Digits_6" )

 hDigits_7 := { 1 => "One",;
 1 => "Two",; // This line replace to previous due to same key 
 3 => "Three",;
 4 => "Four" }
 ListHash( hDigits_7, "Digits_7" )

 * WAIT "EOF digits"

 hCustomers := { "CC001" => "Pierce Firth",;
 "CC002" => "Stellan Taylor",;
 "CC003" => "Chris Cherry",;
 "CC004" => "Amanda Baranski" }
 ListHash( hCustomers, "A hash defined and initialized literally" )
 ?
 ? "Hash value with a specific key (CC003) :", hCustomers[ "CC003" ] // Chris Cherry
 ?
 cKey := "CC003" 
 ?
 ? "Locating a specific record in an hash by key (", cKey, ":"
 ?
 c1Data := hCustomers[ cKey ]
 ? cLMarj, c1Data

 hCountries := { 'Argentina' => "Buenos Aires" }
 hCountries += { 'Brasil' => "Brasilia" }
 hCountries += { 'Chile' => "Santiago" }
 hCountries += { 'Mexico' => "Mexico City" }

 ListHash( hCountries, "A hash defined and initialized by adding with '+=' operator:" )

 hFruits := { "fruits" => { "apple", "chery", "apricot" } }
 hDays := { "days" => { "sunday", "monday" } } 

 hDoris := hFruits + hDays

 ListHash( hDoris, "A hash defined and initialized by concataned two hash with '+' operator:" )

 ?
 @ MAXROW(), 0
 WAIT "EOF HashBasics.prg"

RETURN // HashBasics.Main()
*-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.
PROCEDURE ListHash( hHash, cComment )

 LOCAL x1Pair := NIL

 cComment := IF( HB_ISNIL( cComment ), '', cComment )

 ? 
 ? cComment, "-- Type :", VALTYPE( hHash ), "size:", NTrim ( LEN( hHash ) ) 
 ?
 FOR EACH x1Pair IN hHash
    nIndex := x1Pair:__ENUMINDEX()
    x1Key := x1Pair:__ENUMKEY()
    x1Value := x1Pair:__ENUMVALUE()
    ? cLMarj, NTrim( nIndex ) 
*   ?? '', VALTYPE( x1Pair )
    ?? '', x1Key, "=>"
*   ?? '', VALTYPE( x1Key ) 
*   ?? VALTYPE( x1Value ) 
    IF HB_ISARRAY( x1Value ) 
       AEVAL( x1Value, { | x1 | QQOUT( '', x1 ) } )
    ELSE 
       ?? '', x1Value
    ENDIF 
 NEXT

RETURN // ListHash()
*-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.-._.

HashBass

Array Basics

    An array is a distinct data type which may contains multiple data items under same name.  Data items stored in an array referred as an “element” and can be any data type. An individual element of array referenced by array name and position number of element as an integer in array, called “index” or “subscript”.

 Defining / Building:

    An array is a variable and like all variables has “scope”; arrays can be defined PRIVATE, PUBLIC and LOCAL as well as STATIC.

   Building an array is quite simple: for example to define an array named “aColors” with 5 elements we can use a statement like this:

LOCAL aColors[ 5 ]

or

LOCAL aColors := ARRAY( 5 )

or

LOCAL  aColors := {  , , , , }

Results of these three methods are exactly same; we can inspect easily:

? ValType( aColors )      // A

? LEN( aColors )          // 5
? HB_ValToExp( aColors )  // {NIL, NIL, NIL, NIL, NIL}

NIL is a special data type with meaning “not defined”.

We can define an array with initial value(s)

aColors := {  “green”, ”yellow” ,  “red”,  “black”, “white” }

or assign values after defined:

aColors[ 1 ] := ”green”
aColors[ 2 ] := ”yellow”
aColors[ 3 ] := ”red”
aColors[ 4 ] := ”black”
aColors[ 5 ] := ”white”
? HB_ValToExp( aColors ) // {"green", "yellow", "red", "black", "white"}

Retrieve:

As seen in our first array statement

LOCAL aColors[ 5 ]

used a special sign square brackets as “Array element indicator” ( used also as string delimiter ).

As cited at the beginning, an individual element of array is referenced by array name and position number of element as an integer (enclosed by square brackets) in array, called “index” or “subscript”, and this notation called “subscripting”.

Note that subscribing begins with one.

To specify more than one subscript ( i.e. when using multi-dimensional arrays), you can either enclose each subscript in a separate set of square brackets, or separate the subscripts with commas and enclose the list in square brackets. For example, if aArray is a two dimensional array, the following statements both addresses the second column element of tenth row:

aArray[ 10 ][ 2]
aArray[ 10, 2]

Of course it’s illegal to address an element that is outside of the boundaries of the array (lesser than one or greater than array size / length; one is also illegal for an empty array). Attempting to do so will result in a runtime error.

When making reference to an array element using a subscript, you are actually applying the subscript operator ([]) to an array expression, not only an array identifier (array variable name). An array expression is any valid expression that evaluate to an array. This includes function calls, variable references, subscripting operations, or any other expression that evaluate to an array. For example, the following are all valid:

 { “a”, “b”, “c” }[ 2 ]

x[ 2 ]
ARRAY(3)[ 2 ]
&(<macro expression>)[ 2 ]
(<complex expression>)[ 2 ]

Syntax :

 <aArrayName> [ <nSubscript> ]

<nSubscript> is integer value and indicate sequence number of element into this array.

With this way we can also traverse an array:

FOR nColor := 1 TO LEN( aColors )
     ? aColors[ nColor ]
NEXT nColor

In fact, with the other FOR loop, traversing an array doesn’t require subscripting:

cColor := “” // FOR EACH loop require a predefined loop element

FOR EACH cColor IN aColors
  ? cColor
NEXT nColor

Elements of an array may be different data type; thus arrays called as “ragged” arrays in Clipper language.

aRagged := { 1, "One", DATE(),  .T. } 
FOR nIndex := 1 TO LEN( aRagged )
   x1Elem := aRagged[ nIndex  ]
   ? VALTYPE( x1Elem ), x1Elem
NEXT nIndex

A build-in array function AEVAL() can be use instead of a loop :

AEVAL( aRagged, { | x1 | QOUT(x1 ) } )

 

Adding one element to end of an array:

The architect of array is quite versatile. Array may change ( in size and element values ) dynamically at run time.

 AADD() function can be used for add a new element to the end of an array

AADD(<aTarget>, <expValue>) --> Value

<aTarget> is the array to which a new element is to be added.

<expValue> is the value assigned to the new element.

AADD() is an array function that increases the actual length of the target array by one.  The newly created array element is assigned the value specified by <expValue>.

AADD() is used to dynamically grow an array.  It is useful for building dynamic lists or queues.

For example an array may build empty and later add element(s) to it :

aColors := {}
AADD( aColors, ”green” )
AADD( aColors, ”yellow” )
AADD( aColors, ”red” )
AADD( aColors, ”black” )
AADD( aColors, ”white” )
? HB_ValToExp( aColors ) // {"green", "yellow", "red", "black", "white"}
Inserting one element to an array:

AINS() function can be use for insert a NIL element into an array

AINS (<aTarget>, <nPosition>) --> aTarget

<aTarget> is the array into which a new element will be inserted.

<nPosition> is the position at which the new element will be  inserted.

AINS() is an array function that inserts a new element into a specified array.  The newly inserted element is NIL data type until a new value is assigned to it.  After the insertion, the last element in the array is discarded, and all elements after the new element are shifted down one position.

For a lossless AINS() ( HL_AINS() ) look at attached .prg.

Deleting one element from an array:

ADEL(<aTarget>, <nPosition>) --> aTarget

ADEL() is an array function that deletes an element from an array.  The content of the specified array element is lost, and all elements from that position to the end of the array are shifted up one element.  The last element in the array becomes NIL. So, ADEL() function doesn’t change size of array.

For an another ADEL()  ( HL_ADEL() ) look at attached .prg.

Resizing:

 ASIZE() function can be use for grow or shrink, that is changing size of an array.

ASIZE( <aTarget>, <nLength>) --> aTarget

<aTarget> is the array to grow or shrink.

<nLength> is the new size of the array.

 

     ASIZE() is an array function that changes the actual length of the   <aTarget> array.  The array is shortened or lengthened to match the specified length.  If the array is shortened, elements at the end of the array are lost.  If the array is lengthened, new elements are added to the end of the array and assigned NIL.

     ASIZE() is similar to AADD() which adds a single new element to the end   of an array and optionally assigns a new value at the same time.  Note that ASIZE() is different from AINS() and ADEL(), which do not actually      change the array’s length.

Assigning a fixed value to all elements of an array:

Changing values of all element of an array can not be accomplish by a simple assign statement. For example:

aTest := ARRAY( 3 )
aTest := 1

change type of aTest from array to numeric with a value 1.

Instead, AFILL() function gives a short way to fill an array with a fixed value.

AFILL() :  Fill an array with a specified value

Syntax :

AFILL(<aTarget>, <expValue>,
    [<nStart>], [<nCount>]) --> aTarget
    <aTarget> is the array to be filled.

<expValue> is the value to be placed in each array element.  It can be an expression of any valid data type.

<nStart> is the position of the first element to be filled.  If this argument is omitted, the default value is one.

<nCount> is the number of elements to be filled starting with  element <nStart>.  If this argument is omitted, elements are filled from the starting element position to the end of the array.

Code evaluation on an array:

AEVAL()

Execute a code block for each element in an array

Syntax:

AEVAL(<aArray>, <bBlock>,
    [<nStart>], [<nCount>]) --> aArray

Arguments:

<aArray> is the array to traverse.

<bBlock> is a code block to execute for each element encountered.

<nStart> is the starting element.  If not specified, the default is  element one.

<nCount> is the number of elements to process from <nStart>.  If not specified, the default is all elements to the end of the array.

Returns:

     AEVAL() returns a reference to <aArray>.

Description:

AEVAL() is an array function that evaluates a code block once for each element of an array, passing the element value and the element index as  block parameters.  The return value of the block is ignored.  All      elements in <aArray> are processed unless either the <nStart> or the  <nCount> argument is specified.

AEVAL() makes no assumptions about the contents of the array elements it is passing to the block.  It is assumed that the supplied block knows  what type of data will be in each element.

AEVAL() is similar to DBEVAL() which applies a block to each record of a  database file.  Like DBEVAL(), AEVAL() can be used as a primitive for  the construction of iteration commands for both simple and complex array  structures.

Refer to the Code Blocks section in the “Basic Concepts” chapter of the   Programming and Utilities Guide for more information on the theory and syntax of code blocks.

Examples :

This example uses AEVAL() to display an array of file names  and file sizes returned from the DIRECTORY() function:

#include “Directry.ch”

//

LOCAL aFiles := DIRECTORY(“*.dbf”), nTotal := 0

AEVAL(aFiles, { | aDbfFile |;
     QOUT( PADR(aDbfFile[F_NAME], 10), aDbfFile[F_SIZE]),;
     nTotal += aDbfFile[F_SIZE])} )
//
?
? "Total Bytes:", nTotal

This example uses AEVAL() to build a list consisting of  selected items from a multi-dimensional array:

#include "Directry.ch"
//
LOCAL aFiles := DIRECTORY("*.dbf"), aNames := {}
AEVAL(aFiles, { | file | AADD(aNames, file[F_NAME]) } )

This example changes the contents of the array element depending on a condition.  Notice the use of the codeblock  parameters:

 

LOCAL aArray[6]
AFILL(aArray,"old")
AEVAL(aArray,;
{|cValue,nIndex| IF( cValue == "old",;
aArray[nIndex] := "new",)})

Searching a value into an array :

ASCAN() : Scan an array for a value or until a block returns true (.T.)

Syntax:

ASCAN(<aTarget>, <expSearch>,
       [<nStart>], [<nCount>]) --> nStoppedAt

 Arguments:

<aTarget> is the array to be scanned.

<expSearch> is either a simple value to scan for, or a code block.   If <expSearch> is a simple value it can be character, date, logical, or  numeric type.

<nStart> is the starting element of the scan.  If this argument is not specified, the default starting position is one.

<nCount> is the number of elements to scan from the starting  position.  If this argument is not specified, all elements from the  starting element to the end of the array are scanned.

 Returns:

ASCAN() returns a numeric value representing the array position of the last element scanned.  If <expSearch> is a simple value, ASCAN() returns the position of the first matching element, or zero if a match is not  found.  If <expSearch> is a code block, ASCAN() returns the position of the element where the block returned true (.T.).

 Description:

ASCAN() is an array function that scans an array for a specified value and operates like SEEK when searching for a simple value.  The  <expSearch> value is compared to the target array element beginning with  the leftmost character in the target element and proceeding until there are no more characters left in <expSearch>.  If there is no match,  ASCAN() proceeds to the next element in the array.

Since ASCAN() uses the equal operator (=) for comparisons, it is  sensitive to the status of EXACT.  If EXACT is ON, the target array  element must be exactly equal to the result of <expSearch> to match.

If the <expSearch> argument is a code block, ASCAN() scans the <aTarget>   array executing the block for each element accessed.  As each element is encountered, ASCAN() passes the element’s value as an argument to the code block, and then performs an EVAL() on the block.  The scanning operation stops when the code block returns true (.T.), or ASCAN()  reaches the last element in the array.

  Examples:

This example demonstrates scanning a three-element array using simple values and a code block as search criteria.  The code block criteria show how to perform a case-insensitive search:

aArray := { "Tom", "Mary", "Sue" }
? ASCAN(aArray, "Mary")               // Result: 2
? ASCAN(aArray, "mary")               // Result: 0
//
? ASCAN(aArray, { |x| UPPER(x) == "MARY" }) // Result: 2

This example demonstrates scanning for multiple instances of a search argument after a match is found:

LOCAL aArray := { "Tom", "Mary", "Sue","Mary" },; 
      nStart := 1
//
// Get last array element position
nAtEnd := LEN(aArray)
DO WHILE (nPos := ASCAN(aArray, "Mary", nStart)) > 0
   ? nPos, aArray[nPos]
   //
   // Get new starting position and test
   // boundary condition
   IF (nStart := ++nPos) > nAtEnd
      EXIT
   ENDIF
ENDDO

This example scans a two-dimensional array using a code block.  Note that the parameter aVal in the code block is an array:

LOCAL aArr:={}
CLS
AADD(aArr,{"one","two"})
AADD(aArr,{"three","four"})
AADD(aArr,{"five","six"})
? ASCAN(aArr, {|aVal| aVal[2] == "four"})         // Returns 2

Sorting an array:

ASORT() :  Sort an array

Syntax:

ASORT(<aTarget>, [<nStart>],
[<nCount>], [<bOrder>]) --> aTarget

 Arguments:

<aTarget> is the array to be sorted.

<nStart> is the first element of the sort.  If not specified, the default starting position is one.

<nCount> is the number of elements to be sorted.  If not specified, all elements in the array beginning with the starting element are sorted.

<bOrder> is an optional code block used to determine sorting order.  If not specified, the default order is ascending.

 Returns:

ASORT() returns a reference to the <aTarget> array.

 Description:

 ASORT() is an array function that sorts all or part of an array  containing elements of a single data type.  Data types that can be  sorted include character, date, logical, and numeric.

If the <bOrder> argument is not specified, the default order is ascending.  Elements with low values are sorted toward the top of the  array (first element), while elements with high values are sorted toward the bottom of the array (last element).

If the <bOrder> block argument is specified, it is used to determine the sorting order.  Each time the block is evaluated; two elements from the target array are passed as block parameters.  The block must return true     (.T.) if the elements are in sorted order.  This facility can be used to create a descending or dictionary order sort.  See the examples below.

When sorted, character strings are ordered in ASCII sequence; logical values are sorted with false (.F.) as the low value; date values are sorted chronologically; and numeric values are sorted by magnitude.

 Notes:

ASORT() is only guaranteed to produce sorted output (as defined by the block), not to preserve any existing natural order in the process.

Because multidimensional arrays are implemented by nesting sub-arrays within other arrays, ASORT() will not directly sort  a multidimensional array.  To sort a nested array, you must supply a code block which properly handles the sub-arrays.

 Examples:

This example creates an array of five unsorted elements, sorts the array in ascending order, then sorts the array in descending  order using a code block:

aArray := { 3, 5, 1, 2, 4 }
ASORT(aArray)
// Result: { 1, 2, 3, 4, 5 }
ASORT(aArray,,, { |x, y| x > y })
// Result: { 5, 4, 3, 2, 1 }

This example sorts an array of character strings in ascending order, independent of case.  It does this by using a code block that  converts the elements to uppercase before they are compared:

aArray := { "Fred", Kate", "ALVIN", "friend" }
ASORT(aArray,,, { |x, y| UPPER(x) < UPPER(y) })

This example sorts a nested array using the second element of each sub-array:

aKids := { {"Mary", 14}, {"Joe", 23}, {"Art", 16} }
aSortKids := ASORT(aKids,,, { |x, y| x[2] < y[2] })

Result:

{ {“Mary”, 14}, {“Art”, 16}, {“Joe”, 23} }

Last element in an array:

ATAIL() : Return the highest numbered element of an array

Syntax:

      ATAIL(<aArray>) --> Element

Arguments:

<aArray> is the array.

 Returns:

ATAIL() returns either a value or a reference to an array or object.  The array is not changed.

 Description:

ATAIL() is an array function that returns the highest numbered element  of an array.  It can be used in applications as shorthand for <aArray>[LEN(<aArray>)] when you need to obtain the last element of an      array.

Examples:

The following example creates a literal array and returns that last element of the array:

     aArray := {"a", "b", "c", "d"}
     ? ATAIL(aArray)                     // Result: d

Getting directory info:

ADIR() is a array function to obtain directory information. But it’s  a compatibility function and therefore not recommended. It is superseded by the DIRECTORY() function which returns all file information in a multidimensional array.

A sample .prg : ArrayBasics.prg 

 

ArrBass

Operator overloading

/*
Operator overloading

 Some operators overloaded by extending their functionalities.
"$" was an operator for "checking substring existence in a string" 

 For example :

 ? "A" $ "ABC" // Result: .T.
 ? "Z" $ "ABC" // Result: .F.

 Now, this operator can be used for arrays and hashs too, not only strings.

 See examples below.

 "=>" was a preprocessor operator with meaning "translate to : ...".

Now, this operator can be used as a <key> - <value> separator in Hashs
for define and / or assign <key> - <value> to Hashs.
See examples below.

 "[ ]" was Array element indicator (Special)
 "{ }" was Literal array and code block delimiters (Special)

Now, this indicators can be used for hashs too. 
See examples below.

"+=" is self-increment operator that can be used both numeric 
and string values.

 Such as :

 cTest := "This"
 cTest += " is" 
 ? cTest // This is

 nTest := 3
 nTest += 10
 ? nTest // 13

 Now, this operator can be used for adding elements to an existing hash;
 ( but no for arrays ! ).
 Note : Extended functionalities of $ and += operators depends xHB lib.
        So need this usages to xHB lib and xHB.ch.

 See examples below.

*/
#include "xhb.ch"
PROCEDURE Main()

 CLS

 aFruits := { "apple", "appricot", "cherry", "melon", "pear", "mulberry" }

 ? "aFruits", IF( "pear" $ aFruits, '', 'not ' ) + "contain pear"
 ? "aFruits", IF( "grapes" $ aFruits, '', 'not ' ) +"contain grapes"

 aComplex := ARRAY( 10 )
AEVAL( aComplex, { | x1, i1 | aComplex[ i1 ] := i1 } )

 aComplex[ 5 ] := DATE()
 aComplex[ 7 ] := .F.

 ?
 ? "aComplex", IF( 3 $ aComplex, '', 'not ' ) + "contain 3"
 ? "aComplex", IF( 13 $ aComplex, '', 'not ' ) + "contain 13"
 ? "aComplex", IF( .T. $ aComplex, '', 'not ' ) + "contain .T."
 ? "aComplex", IF( .F. $ aComplex, '', 'not ' ) + "contain .F."
hEmpty := { => }
 ?
 ? "hEmpty is a", VALTYPE( hEmpty ), "type variable have",;
 STR( LEN( hEmpty ), 1 ), "element and it's",;
 IF( EMPTY( hEmpty ), '', 'not' ), "Empty"
hCountries := { 'Argentina' => "Buenos Aires" }
 hCountries += { 'Brasil' => "Brasilia" }
 hCountries += { 'Chile' => "Santiago" }
 hCountries += { 'Mexico' => "Mexico City" }

 ?
 ? "hCountries is a", VALTYPE( hCountries ), "type variable have",;
 STR( LEN( hCountries ), 1 ), "elements and and it's",;
 IF( EMPTY( hCountries ), '', 'not' ), "Empty"
cCountry := NIL
 FOR EACH cCountry IN hCountries
 ? cCountry:__ENUMKEY(), "=>", cCountry:__ENUMVALUE()
 NEXT 

 hDays := { 'Days' => { "Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun" } }

 ?
 ? "hDays", IF( 'Days' $ hDays, '', 'not ' ) + "contain Days" 
 ? "hDays", IF( "Mon" $ hDays, '', 'not ' ) + "contain Mon" 
 ? "hDays['Days']", IF( "Fri" $ hDays["Days"], '', 'not ' ) + "contain Fri"
hLanguages := { "EN" => "English" } +; 
 { "DE" => "Deutsche" } +; 
 { "ES" => "Español" } +; 
 { "FR" => "Français" } +; 
 { "IT" => "Italiano" } +; 
 { "PL" => "Polkski" } +; 
 { "PT" => "Português" } +; 
 { "RU" => "Russkî" } +; 
 { "TR" => "Türkçe" }

 ?
 ? "hLanguages is a", VALTYPE( hLanguages ), "type variable have",;
 STR( LEN( hLanguages ), 1 ), "elements and and it's",;
 IF( EMPTY( hLanguages ), '', 'not' ), "Empty"
cLanguage := NIL 
 FOR EACH cLanguage IN hLanguages
 ? cLanguage:__ENUMKEY(), "=>", cLanguage:__ENUMVALUE()
 NEXT 

 @ MAXROW(), 0 
 WAIT "EOF OprOLoad.prg"
RETURN // OprOLoad.Prg.Main()
OprOLoad

Using code blocks, again

Using code blocks again (.pdf)

Array and GET tricks

Array and GET tricks

Simple Clipper Extensions

Simple extensions

Exact comparisons,
Name precedence,
SELECT 0,
Alias functions,
Call by reference and value,
Other Clipper extensions