Programmer 7500

home *** CD-ROM | disk | FTP | other *** search

/ Programmer 7500 / MAX_PROGRAMMERS.iso / CLIPPER / DATABASE / BCSMEM.ZIP / ARRAY.DOC next >

Wrap

Text File | 1993-05-06 | 38.4 KB | 914 lines

(10U Clipper Multi-dimensional Arrays by Roger Donnay A. Introduction B. The Evolution of the Multi-dimensional array C. Classes of Multi-dimensional arrays 1. Parallel symmetrical arrays 2. Ragged symmetrical arrays 3. Asymmetrical arrays D. Working with Linked arrays E. Browsing a Multi-dimensional array F. Saving and Restoring Multi-dimensional arrays ╔════════════════╗ ║ INTRODUCTION ║ ╚════════════════╝ In my opinion, the single most powerful new feature of Clipper 5.0 is the multi-dimensional array system. Sometimes referred to as "ragged arrays", this incredibly versatile data type is almost unlimited in capability and opens the door to a new kind of programming just not possible in Summer 87. This seminar explores the use of ragged arrays from a practical and theoretical viewpoint and is directed to the Clipper programmer who is still discovering the power of Clipper 5.0. I don't usually like to use a lot of code snippets in my seminars, but teaching multi-dimensional arrays is simplified by using lots of example code, so feel free to use the code provided. Much of the code examples are extracted from the dCLIP libraries. ╔═══════════════════════════════════════════════╗ ║ THE EVOLUTION OF THE MULTIDIMENSIONAL ARRAY ║ ╚═══════════════════════════════════════════════╝ Clipper Summer 87 supported "single-dimensional" arrays. The data structure allowed the programmer to store information in a structure called data "elements". Most languages which support arrays must first pre-define the length of the array (number of elements) and the length and type of each element of the array. This is because memory must be pre-allocated to allow for storage of data in the array elements. This memory allocation at compile/link time is a type of "early binding" to insure that memory will always be available during the running of the application. Clipper Summer 87 attempted to improve on this concept by eliminating the need to pre-define the type and size of each element of the array so the programmer could change any element at will. This new kind of "dynamic array" made programming much easier and more powerful, but the programmer now needed to consider the effects that this dynamic memory allocation may have on the use of the memory pool at run-time. Dynamic arrays that were not properly managed could easily run the program out of memory. Fortunately, this did not occur often in Summer 87 because even though any element of the array could be accessed via one (1) symbol declaration (the array name), the memory allocation and de-allocation for the elements was handled on an individual basis. Example of a single-dimensional array (S87 code): * make an array of 5 elements DECLARE myArray[5] * initialize each element as a null string value AFILL( myArray, "" ) * make element 2 a date value myArray[2] = DATE() * make element 4 a numeric value myArray[4] = 234.56 The array (myArray) will now look like this: myArray[1] = "" myArray[2] = 10/12/92 myArray[3] = "" myArray[4] = 234.56 myArray[5] = "" Probably the single most significant advantage to using arrays in Summer 87 was that any element of an array could be used in expressions exactly like any memory variable. Another significant advantage, but maybe not so obvious, was how using arrays instead of memvars reduced the memory requirements of the application. This is because an entire array uses only one symbol, whereas each memory variable uses a separate symbol. -- Why multi-dimensional arrays? -- Arrays in Summer 87 made the programmer's life much simpler. Scatter-gather routines, get lists, pick-lists, browse-screens, menus, etc. became easier to program using arrays because the size of the array and each element could be determined during the actual running of the application. Arrays, like databases, are most effectively used and maintained when they are normalized. "Normalizing" a database is simply a term describing the careful structuring of a set of relational databases so that information can be stored in related groups (ie, invoices, customers, etc.) with little or no data redundancy. Good programming practices require that arrays also be structure in sets of related information. For example, creating an array-driven browse-system would require the use of many arrays for storing information like (1) screen-coordinates, (2) field-lists, (3) column-widths, (4) data-buffers, (5) relations, etc..etc. Summer 87 provided the ability to create many different arrays, but no easy mechanism to further relate these sets of arrays to each other. In the above example, each array had to be given a separate name. If you wished to save and restore a complete browse configuration, the process was difficult and impractical and required that the information be stored and retrieved from databases or kept alive as PUBLIC arrays and symbols. I attempted, with a good degree of success, to eliminate the separately defined arrays in my browse system and tie all the groups of information together into a single array which was a "psuedo" multi-dimensional array. I found I could eliminate the separate arrays and group everything together into one array by maintaining a group of offset pointers into one long array. For example, elements 1-4 would be the screen coordinates, 5-55 the field names, 56-106 the column order, etc. A sample of code that accessed an element of this "psuedo" multi-dimensional array would look like this: fieldname = aBrowse[ columns[ columnOffset + currentcol ] ] ) A psuedo-multi-dimensional array would be treated like this: DECLARE myArray [ 4 + fcount()*2 ] fieldoffset = 4 columnoffset = 4 + fieldoffset * put default screen coordinates into array myArray [1] = 0 myArray [2] = 0 myArray [3] = 24 myArray [4] = 79 * fill field name area of array for i = 1 to fcount() myArray[ fieldoffset + i ] = field(i) next * fill column pointer area of array with default column widths for i = 1 to fcount() myArray[ columnoffset + i ] = len(transform(field(i),'@') next Psuedo-multi-dimensional arrays in S87 were possible, but soon the code started becoming difficult to manage with every line of code making some reference to an array element "number" and offset "number" rather than an easy-to-remember symbol "name". Furthermore, it was impossible to dynamically re-size any sub-array without completely re-building the entire array. These limitations made it impractical to do multiple-window browsing, dynamic adding and deleting of columns, one-to-many browses, etc. Regardless of the limitations of the Summer 87 array system, it was still so powerful that Clipper applications grew in size and functionality at an alarming rate, thereby creating a whole new set of problems. The average Clipper programmer didn't want to invest in a deeper understanding of the memory ramifications of the heavy use of arrays, so he or she would work around the problem by using third-party DOS memory managers and overlay linkers that would give the application more DOS memory to start with. These third-party products were wonderful and extended the lifetime of S87, but eventually these applications will need to be upgraded to a language with a better memory manager. -- Clipper 5.0 arrays -- Clipper 5.0 looks at arrays in an entirely new light. To overcome the difficulties that caused us to out-grow the S87 array system, the 5.0 development team needed to treat arrays not as arrays in the traditional sense but as "linked lists". The need to grow arrays, shrink arrays, pass parameters in arrays, return arrays as values, and even store arrays and code-blocks in other arrays required developing a system to manage "pointers" and "pointers to pointers". The ultimate result of this architecture is a system that not only optimizes the use of symbols, but also the use of memory for the data elements. This automatic system of "optimizing" the data elements allows for creating incredibly large arrays that can be passed around the application for use by many routines and sub-routines by simply passing pointers rather than data. This concept vastly improves memory and speed performance and revolutionizes the future of data-driven programming. The development team took this concept even further and went as far as managing the pointers to the sub-elements so as to simply create a list of pointers to a "common" memory location for elements that contained identical data. For example, in Summer 87 or just about any other language, if you create an array of 1000 elements, and initialize each element to a length of 100 characters, a memory block of 100,000 (100*1000) bytes would be required. Clipper 5.0 effectively handles this task by creating a system of logical pointers to the same area of memory until the information in an element is modified. This dynamic optimization will cause even a huge array to allocate very little memory when it is created, and from then on, only the absolute minimum amount needed as the array information changes. This concept, by itself, is a significant advancement in array architecture and memory management, but the development team went even further and created a Virtual Memory Manager (VMM) system that stores the array data in EMS memory. To see how effective this array optimization system really is, try this little test: for i := 1 to 25 arrayname = 'array'+ALLTRIM(STR(i)) &arrayname = array(1000) afill(&arrayname,space(1000)) ? arrayname, memory(0), memory(4) next The above code will create 25 arrays of 1000 elements each and and 100 characters in each element. This is theoretically a memory requirement of 25 megabytes. In actuality the above code when run under dCLIP uses 0 bytes of free pool memory ( MEMORY(0) ) and only 340k of VMM ( MEMORY(4) ). ╔══════════════════════════════════════╗ ║ CLASSES OF MULTIDIMENSIONAL ARRAYS ║ ╚══════════════════════════════════════╝ Clipper 5.0 provides a variety of methods to create and use multi-dimensional arrays. I am going to attempt to create a classification for types of multi-dimensional arrays. These classes are provided only to help define the different types of array programming and the advantages or disadvantages of each. Clipper does not make any distinction and treats all arrays the same way. 1. Parallel symmetrical arrays. Parallel arrays are the most common arrays in Clipper and are usually created using the ARRAY() function to predefine the length of the array. Parallel is a term defining the way data elements in the array are "grouped". An example of parallel arrays would be an array of 3 sub-arrays in which each sub-array has the same number of elements, and each element in each sub-array has a direct relationship with the corresponding element in every other sub-array. For example, a parallel array containing the structure of a database with 4 fields would look like this: Type Contents aStru[1] A Sub-array of field names aStru[1,1] C CUST_NAME aStru[1,2] C SHIP_DATE aStru[1,3] C BALANCE aStru[1,4] C PRINT_FLAG aStru[2] A Sub-array of field types aStru[2,1] C C aStru[2,2] C D aStru[2,3] C N aStru[2,4] C L aStru[3] A Sub-array of field lengths aStru[3,1] C 25 aStru[3,2] C 8 aStru[3,3] C 9 aStru[3,4] C 1 aStru[4] A Sub-array of field decimals aStru[4,1] N 0 aStru[4,2] N 0 aStru[4,3] N 2 aStru[4,4] N 0 This array could be created like this: aStru := Array( 4, Fcount() ) FOR i := 1 TO Fcount() aStru[1,i] := Field(i) aStru[2,i] := Type(Field(i)) aStru[3,i] := Len(Transform(&(aStru[1,i]),'@')) aStru[4,i] := Eval( {|n| n := At('.', ; Transform(&(astru[1,i]),'@')), ; Iif( n > 0, astru[3,i] - n , 0 ) } ) NEXT Parallel arrays were very common in Summer 87 code, except that each array was assigned a different name or symbol, whereas parallel arrays in 5.0 can all exist under one array name and are addressed by sub-element number. Parallel arrays are most commonly used when you need a pick-list of items (using ACHOICE() or similar function) and then need to index into a set of parallel arrays to extract all related information based on a pointer returned by picking an item from the first parallel array. 2. Ragged symmetrical arrays. Ragged symmetrical arrays are a new concept introduced in Clipper 5.0 and are structured in such a way as to make the data in the array elements much easier to evaluate in a single expression with the AEVAL() function. An example of ragged symmetrical arrays would be an array of 4 sub-arrays in which each sub-array has the same number of elements (i.e. the same "structure") however, there is no direct relationship between the information in any one sub-array and any other sub-array. Instead, all the pertinent related information is contained in all the elements of a single sub-array. Let's create a ragged symmetrical array containing the structure of the same database we used in the parallel array example above. The array would look like this: Type Contents aStru[1] A Field 1 Information aStru[1,1] C CUST_NAME aStru[1,2] C C aStru[1,3] C 25 aStru[1,4] N 0 aStru[2] A Field 2 Information aStru[2,1] C SHIP_DATE aStru[2,2] C D aStru[2,3] C 8 aStru[2,4] N 0 aStru[3] A Field 3 Information aStru[3,1] C BALANCE aStru[3,2] C N aStru[3,3] C 9 aStru[3,4] N 2 aStru[4] A Field 4 Information aStru[4,1] C PRINT_FLAG aStru[4,2] C L aStru[4,3] C 1 aStru[4, ] N 0 This array could be created like this: aStru := {} FOR i := 1 TO Fcount() Aadd( aStru, Array(4) ) aStru[i,1] := Field(i) aStru[i,2] := Type(Field(i)) aStru[i,3] := Len(Transform(&(aStru[i,1]),'@')) aStru[i,4] := Eval( {|n| n := At('.', ; Transform(&(astru[i,1]),'@')), ; Iif( n > 0, astru[i,3] - n , 0 ) } ) NEXT The main advantage of organizing your arrays in ragged symmetrical form is in the ease of evaluating the information in the array. For example, listing the structure of a database that has been loaded into a ragged symmetrical array would require only a single expression: Aeval( aStru, { |a| qout(pad(a[1],a[2],a[3],a[4]) } ) Whereas trying to list the same information from parallel arrays would require more complex structural code: FOR i := 1 TO Fcount() qout(pad(aStru[1,i],10),aStru[2,i],aStru[3,i],aStru[4,i]) NEXT The disadvantage, however of organizing data in ragged symmetrical form is that the array cannot be simply passed to a pick-list function such as ACHOICE(), so here's a handy function to convert one array type to another. FUNCTION dc_aconvert ( aInput ) LOCAL i, j, aOutput := {} aOutput := Array( Len( aInput[1] ), Len( aInput ) ) FOR i := 1 TO LEN( aInput ) FOR j := 1 TO LEN( aInput[1] ) aOutput[j,i] := aInput[i,j] NEXT NEXT RETURN aOutput You would use this function as follows: aRaggedDir := Directory() aParallelDir := DC_ACONVERT( aRaggedDir ) nChoice := Achoice( 10,10,20,40, aParallelDir[1] ) ? 'The size of '+aParallelDir[1,nChoice]+' is '+ ; Str( aParallelDir[2,nChoice] ) 3. Asymmetrical arrays. An asymmetrical array is simply an array of information in which no element of the array has a "direct" indexed or ordinal relationship with any other element of the array yet all the information in all the elements make up a complete package of information. Basically, asymmetrical arrays can be used in place of memvars, in which each element of the array represents a specific piece of information. So why not use memvars instead of arrays? Mainly, because memvars cannot be easily grouped together into a complete "set" of information that can be easily stored and retrieved, whereas and entire array can be saved, restored, or even passed as a single parameter. An example of an asymmetrical array would be an array in which element 1 is another array which always represents the screen coordinates of a browse screen, element 2 is a character field with color of the screen, element 3 is a numeric field representing the work area, element 4 is a logical field, element 5 is a date field, etc., etc. In fact an asymmetrical array may even contain sub-arrays that are ragged symmetrical or parallel symmetrical. This entire multi-dimensional array could contain all the information necessary to repaint all browse screens for all work areas. This kind of array is effectively an "object", because it encapsulates all the necessary information about a program configuration into a nice, neat package. The advantage of working with large asymmetrical arrays is that only one symbol is used for the entire array, making it easy to save, restore and pass around an application. The disadvantage, is that the programmer must remember which element of which sub-array contains the desired data thereby making the source code very cryptic, unreadable, and difficult to maintain. In addition, if it became necessary to add new elements to the array, the ordinal position of each other element may change. This is where the pre-processor is not only very handy but an absolute necessity when working with large asymmetrical arrays. Assigning a "manifest constant" to an array element now allows the programmer to use multi- dimensional array elements and sub-elements in the same way he/she uses memory variables. It isn't necessary to remember which sub-element of an array contains a piece of information when a #define statement can be used to create any name desired. From then on, the programmer can write or read array information via an easy-to-remember set of variable names. There are several schools of thought when #defining manifest constants to represents array elements. Many programmers prefer to use the #define "only" for defining numeric values. Here is an example of this kind of array pre-processing: // BROWSE sub-array definitions #define MAIN 1 #define COORDINATES 2 #define FIELDS 4 // COORDINATES definitions #define STARTROW 1 #define STARTCOL 2 #define ENDROW 3 #define ENDCOL 4 Defining array manifests in this manner requires that you use more than 1 variable name to access sub-array information. For example you would get the value of the start row of the display like this: nStartRow := BROWSE[ nWorkArea, COORDINATES, STARTROW ] Another method of #defining array pointers is by using one a "manifest symbol" rather than a "manifest constant". Here is an example of manifest symbols: // BROWSE sub-array definitions #define aMAIN 1 #define aCOORDINATES 2 #define aFIELDS 4 // COORDINATES definitions #define nSTART_ROW BROWSE[ nWorkarea, 2, 1 ] #define nSTART_COLUMN BROWSE[ nWorkarea, 2, 2 ] #define nEND_ROW BROWSE[ nWorkarea, 2, 3 ] #define nEND_COLUMN BROWSE[ nWorkarea, 2, 4 ] Defining array manifests in this manner allows you to access multi-dimensional array information via shorter and simpler commands. For example you would get the value of the start row of the display like this: nStartRow := nSTART_ROW There currently is no "Hungarian notation" standard for manifest constants so you can use any name you wish, but I prefer the following format: xYYYYYYYYYY - -------- | | | -------- The assigned name (in all capitals) ------------- The "type" of the data (in lower case) (a-Array, n-Numeric, d-Date, c-Character, etc) Defining manifest contstants in this manner makes them stand out in your program and not get confused with your memory variables or database field names. ╔══════════════════════════════╗ ║ WORKING WITH LINKED ARRAYS ║ ╚══════════════════════════════╝ Although array elements in Clipper 5.0 can store the same types of data as memvars, it must be clearly understood that the 5.0 linked array system creates new "pointers", NOT new "values". If this truth is forgotten or misunderstood you may not get the expected results in your program. Let's take the following code as an example: n1 := 1 n2 := n1 n2 := 2 ? n1 1 a1 := { 1, 2, 3 } a2 := a1 a2[1] := 10 ? a1[1] 10 In the numeric memvar example above, n2 := n1 assigns the "value" of n1 to n2. There is no other relationship between n1 and n2. The value of n2 can now be changed to anything without affecting the value of n1. In the array example above, a2 := a1 assigns the "address" of a1 to a2. This now creates a direct relationship between a1 and a2 by creating a new set of pointers to the same information. Now if any of the elements of a2 are changed, the information in a1 will also be changed. If it desired to copy the information from an array to a new array and then work with the new array without affecting the original array, use the ACLONE() function. Actually, this "linked-list" type of array system is not a disadvantage, but in the contrary, is the reason why arrays in Clipper 5.0 are so powerful. When an array is passed as a parameter to another function or returned as a value from a function, only the pointer to the same information in memory is actually passed, not a set of values. This means that if the data in a LOCAL array that has been passed to a function is modified, the original passed array is also modified. Look at the following code: PROCEDURE TEST LOCAL a1 := {1,2,3} LOCAL a2 a2 := TEST2 ( a1 ) ? a1[1] 10 STATIC FUNCTION TEST2 ( a3 ) LOCAL a4 := a3 a4[1] := 10 a4[2] := 11 a4[3] := 12 RETURN a4 In the above example, even though a4 in function TEST2 has been designated as a LOCAL memvar, the assigment of a3 to a4 will force the information in a1 of procedure TEST to change when it is changed in function TEST2. ╔═════════════════════════════════════╗ ║ BROWSING A MULTIDIMENSIONAL ARRAY ║ ╚═════════════════════════════════════╝ I have seen lots of array browsers that don't give me the dimensional view of the array that I like. Most array browsers are written using Tbrowse, this one is not. This browser not only views the information in an array or object but also allows you to change the value and type of any array element or sub- element, or to grow or shrink any sub-array. Browsing an array in this manner requires building a different image of the array using macros and a Private memvar. This may not be a desirable programming practice to some, but it yields some good results. Note: This array browser can take a few seconds to build the Achoice image of the array if it is a large array. STATIC nLastKey := 0 MEMVAR aNewArray FUNCTION abrowse ( aArray ) LOCAL aArrayView, nSelect, nStart, cSaveScreen, cSaveScrn2,; cArrayElem, xValue, cType, nValueLoc, nRow, cElement, nLength,; cValue, cOldType, nOldLength, cSubName, nElement, lReBuild,; nLastElem, getlist := {} IF !(VALTYPE(aArray)$'AO') RETURN .f. ENDIF PRIVATE aNewArray := aArray aArrayView := ABROWSE3( {}, aNewArray, '', 0 ) @ 0,2 TO 24,79 DOUBLE @ 24,7 SAY "┤ ENTER = Edit, ESCape = Exit, + = Add, "+; "DEL = Delete, INS = Insert ├" nSelect := 1 nStart := 1 DO WHILE .t. SET KEY 7 TO ABROWSE4 SET KEY 22 TO ABROWSE4 SET KEY 43 TO ABROWSE4 nLastKey := 0 nSelect := ACHOICE(1,3,23,78,aArrayView,,,nSelect,nStart) SET KEY 7 TO SET KEY 22 TO SET KEY 43 TO nStart := ROW()-1 IF LASTKEY()=27 EXIT ENDIF IF nSelect = 0 LOOP ENDIF cArrayElem := aArrayView[nSelect] cType := SUBSTR( cArrayElem, AT( '],', cArrayElem)+2, 1 ) nValueLoc := AT( '],', cArrayElem ) + 4 cValue := SUBSTR( cArrayElem, nValueLoc ) cElement := SUBSTR( cArrayElem, 1, nValueLoc-4 ) nRow := IIF( ROW()>11, 2, 12 ) cArrayElem := 'aNewArray' + cElement nLength := LEN(IIF(cType#'A',cValue,&cArrayElem)) lReBuild := .f. FOR nLastElem := LEN(cElement) TO 1 STEP -1 IF SUBSTR(cElement,nLastElem,1) $ '[,' EXIT ENDIF NEXT cSubName := 'aNewArray' + ALLTRIM(SUBSTR(cElement,1,nLastElem-1))+; IIF(SUBSTR(cElement,nLastElem,1)=',',']','') nElement := VAL(STRTRAN(SUBSTR(cElement,nLastElem+1),']','')) nOldLength := nLength DO CASE CASE nLastKey = 22 // Insert key inserts an element AINS( &cSubName, nElement) lReBuild := .t. CASE nLastKey = 7 // Delete key deletes an element ADEL( &cSubName, nElement) lReBuild := .t. CASE nLastKey = 43 // + key adds a new element AADD( &cSubName, nil ) lReBuild := .t. CASE LASTKEY()=13 cSaveScrn2 := SaveScreen( nRow,5,nRow+4,75 ) @ nRow,5 CLEAR TO nRow+4,75 @ nRow,5 TO nRow+4,75 cOldType := cType @ nRow+1,7 SAY ' Type' GET cType PICT '!' VALID cType$'ACDNLU' READ DO CASE CASE cType = 'L' xValue := 'T'$cValue CASE cType = 'D' xValue := CTOD(cValue) CASE cType = 'N' xValue := VAL(cValue) CASE cType = 'U' xValue := nil CASE cType $ 'CA' @ nRow+2,8 SAY 'Length' GET nLength READ IF cType = 'C' xValue := PAD(cValue,nLength) ENDIF ENDCASE DO CASE CASE cType = 'N' @ nRow+3,8 SAY ' Value' GET xValue PICT ; '9999999999999.9999999' READ CASE !cType $ 'UA' @ nRow+3,8 SAY ' Value' GET xValue PICT '@S58' READ ENDCASE IF LASTKEY()#27 IF cType#'A' IF cType='U' aArrayView[nSelect] := cElement + ','+cType+',' ELSE aArrayView[nSelect] := cElement + ','+cType+',' +TRANSFORM(xValue,'@A') ENDIF &cArrayElem := xValue ELSE IF TYPE(cArrayElem)#'A' &cArrayElem := {} ENDIF ASIZE(&cArrayElem,nLength) ENDIF lReBuild := (cOldType # cType .AND. ; ( cType='A' .OR. cOldType='A' )) ; .OR. (cType='A' .AND. cOldType='A' ; .AND. nLength # nOldLength) ENDIF RestScreen(nRow,5,nRow+4,75,cSaveScrn2) ENDCASE IF lReBuild aArrayView := ABROWSE3( {}, aNewArray, '', 0 ) ENDIF ENDDO DC_IMPL(cSaveScreen) RETURN .t. // ----------------------------------------------------------- // STATIC FUNC aBrowse3 ( aArrayView, aArrayDisp, cElement, nRecurs ) LOCAL nElement, cTextLine, cType, nArrayLen, cValue, cElement2 nArrayLen := LEN(aArrayDisp) FOR nElement := 1 TO nArrayLen cType := VALTYPE(aArrayDisp[nElement]) cTextLine := '' cElement2 := cElement+'['+ALLTRIM(STR(nElement,4))+']' cElement2 := STRTRAN( cElement2, '][', ',' ) DO CASE CASE cType='C' cTextLine := aArrayDisp[nElement] CASE cType='N' cTextLine := STR(aArrayDisp[nElement]) CASE cType='D' cTextLine := DTOC(aArrayDisp[nElement]) CASE cType='L' cTextLine := IIF(aArrayDisp[nElement],'T','F') CASE cType$'AO' AADD( aArrayView, SPACE(nRecurs)+cElement2+','+cType+','+; cTextLine ) aArrayView := ABROWSE3( aArrayView, aArrayDisp[nElement], ; cElement2, nRecurs+1 ) LOOP ENDCASE AADD( aArrayView, SPACE(nRecurs)+cElement2+','+cType+','+; cTextLine ) NEXT RETURN aArrayView // ---------------------------------------------------------- // STATIC PROC ABROWSE4 nLastKey := LASTKEY() KEYBOARD CHR(13) RETURN // ---------------------------------------------------------- // ╔═══════════════════════════════╗ ║ SAVING AND RESTORING ARRAYS ║ ╚═══════════════════════════════╝ Saving and restoring arrays to and from a disk file can offer the advantage of creating "persitent" objects. The following code will save and restore complete multi-dimensional arrays that DO NOT contain code blocks. Currently, there is no mechanism provided within Clipper to save/restore code blocks to disk, only to memvars. If you are using code blocks in arrays, you must store the code block as a string and macro- compile the string to restore the code block. The code provided here is 100% clipper code for compatability purposes only. The array saving/restoring mechanism I prefer is identical to the below code, except that the F_EOF and F_READLINE functions are replaced with assembly-level functions for speed. MEMVAR aOutArray FUNCTION arrayread ( cFileName ) LOCAL cTextLine, cType, cElement, nArrayLen, nHandle,; nComma, cValue, cArray PRIVATE aOutArray := {} nHandle := FOPEN( cFileName ) IF nHandle<=0 RETURN nil ENDIF nArrayLen := 0 DO WHILE !F_EOF(nHandle) cTextLine := F_READLINE(nHandle) IF LEFT(cTextLine,1) = '[' .OR. LEFT(cTextLine,1) = ',' nComma := AT(',',cTextLine) cElement := TRIM(SUBSTR(cTextLine,1,nComma-1)) cType := UPPER(SUBSTR(cTextLine,nComma+1,1)) cValue := SUBSTR(cTextLine,nComma+3) cArray := 'aOutArray'+cElement ELSE cValue := cValue + cTextLine ENDIF DO CASE CASE cType='C' &cArray := STRTRAN(cValue,CHR(255),CHR(13)+CHR(10)) CASE cType='N' &cArray := VAL(cValue) CASE cType='D' &cArray := CTOD(cValue) CASE cType='L' &cArray := IIF(cValue='Y',.t.,.f.) CASE cType='A' cArray := 'aOutArray'+cElement &cArray := {} ASIZE(&cArray,VAL(cValue)) CASE cType='B' &cArray := &('"'+cValue+'"') ENDCASE ENDDO FCLOSE(nHandle) RETURN aOutArray // ------------------------------------------------------------ // FUNCTION arraywrite ( aArray, cFileName ) LOCAL cTextLine, cType, nElement, nHandle IF VALTYPE(aArray)#'A' RETURN .F. ENDIF nHandle := FCREATE( cFileName ) IF nHandle<=0 RETURN .f. ENDIF _DCARRAY_W2 ( aArray, nHandle, '' ) FCLOSE(nHandle) RETURN .T. // ------------------------------------------------------------- // STATIC FUNCTION _dcarray_w2 ( aArray , nHandle, cElement ) LOCAL nElement, cTextLine, cType, nArrayLen, cValue nArrayLen := LEN(aArray) FWRITE(nHandle,cElement+',A,'+ALLTRIM(STR(nArrayLen))+; CHR(13)+CHR(10)) FOR nElement := 1 TO nArrayLen IF VALTYPE(aArray[nElement])='U' LOOP ENDIF cValue := aArray[nElement] cType := VALTYPE(cValue) cTextLine := '' DO CASE CASE cType='C' cTextLine := STRTRAN(HARDCR(cValue),CHR(13)+; CHR(10),CHR(255)) CASE cType='N' cTextLine := ALLTRIM(STR(cValue)) CASE cType='D' cTextLine := DTOC(cValue) CASE cType='L' cTextLine := IIF(cValue,'Y','N') CASE cType='A' _DCARRAY_W2 ( aArray[nElement], nHandle, ; cElement+'['+ALLTRIM(STR(nElement,4))+']') LOOP ENDCASE FWRITE(nHandle,cElement+'['+ALLTRIM(STR(nElement,4))+']'; +','+cType+','+cTextLine+CHR(13)+CHR(10)) NEXT RETURN .T. // -------------------------------------------------------- // STATIC FUNCTION f_eof ( nHandle ) LOCAL nCurrent, lEof nCurrent := FSEEK(nHandle,0,1) lEof := .f. IF nCurrent >= FSEEK(nHandle,0,2) lEof := .t. ENDIF FSEEK(nHandle,nCurrent,0) RETURN lEof // --------------------------------------------------------- // STATIC FUNCTION f_readline ( nHandle ) LOCAL cString, nCurrent, nCr nCurrent := FSEEK(nHandle,0,1) cString := FREADSTR(nHandle,500) nCr := AT (CHR(13),cString) FSEEK(nHandle,nCurrent,0) FSEEK(nHandle,nCr+1,1) RETURN SUBSTR(cString,1,nCr-1)