NetNews Usenet Archive 1993 #3

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

/ NetNews Usenet Archive 1993 #3 / NN_1993_3.iso / spool / comp / database / informix / 3041 < prev next >

Wrap

Internet Message Format | 1993-01-23 | 48.8 KB

Path: sparky!uunet!cs.utexas.edu!swrinde!emory!emory!not-for-mail From: obelix.informix.com!johnl@emory.mathcs.emory.edu (Jonathan Leffler) Newsgroups: comp.databases.informix Subject: I4GL Modifications: Part 1 of 3 Date: 22 Jan 1993 23:18:09 -0500 Organization: Mailing List Gateway Lines: 963 Sender: walt@mathcs.emory.edu Distribution: world Message-ID: <1jqgu1INN8di@emory.mathcs.emory.edu> Reply-To: obelix.informix.com!johnl@emory.mathcs.emory.edu (Jonathan Leffler) NNTP-Posting-Host: emory.mathcs.emory.edu X-Informix-List-ID: <list.1813> : "@(#)shar2.c 2.3 92/11/05" #!/bin/sh # shar: Shell Archiver (v1.22) # # This is a shell archive. # Remove everything above this line and run sh on the resulting file # If this archive is complete, you will see this message at the end # "All files extracted" # # Created: Fri Jan 22 19:12:20 1993 by jl at JLSS # # This is part 1 of a multipart archive # do not concatenate these parts, unpack them in order with /bin/sh # Files archived in this archive: # i4glmods # if test -r s2_seq_.tmp then echo "Must unpack archives in sequence!" next=`cat s2_seq_.tmp`; echo "Please unpack part $next next" exit 1; fi echo "x - i4glmods" sed 's/^X//' << 'SHAR_EOF' > i4glmods && X X ********************************************* X ** SUGGESTED MODIFICATIONS TO INFORMIX-4GL ** X ********************************************* X XDisclaimer: X Despite the fact that I work for Informix Software in the UK, this list X of suggestions does not have any more substance than you care to give X it by materialising it on your screen, or by placing it on paper. It X must not be taken as any sort of indication of an official stance by X Informix Software. X XMost of the suggestions in this list have been collated from the recent Xdiscussions on comp.databases.informix and the mailing list run by Walt XHultgen (walt@rmy.emory.edu); a few have been collected from other sources. XI claim no credit for the basic requirements stated below, though the Xstatement below is mainly in my own words. If people desparately want to Xclaim the idea, please send me an abbreviated version of this document with Xthe ideas to which you lay claim suitably annotated. I will try to include Xyour claims in any future version of the document. Bear in mind that more Xthan one person may have made the same suggestion -- in fact, that is Xgenerally the case. X XThis is a long document. To try and help make it manageable, it is split Xinto several sections. The first section is the preamble you're reading Xnow. The second section is a brief statement of the features -- one line Xindicating what is required for each feature. The items in this section Xare rated for plausibility with a letter in the set A-F with A good and F Xbad. An A-rated idea generally fits closely with the existing language, Xshould be easy to implement, and would be widely used. An F-rated idea Xprobably does violence to the spirit of the existing language, would be Xhard to implement, and might not be widely used. This scale is completely Xsubjective. The third section expands on each of the ideas in the second Xsection. Sometimes it proposes chunk of language syntax; this should be Xtaken as a suggestion, and you should consider sending in a revised Xversion. Whenever you do this, please bear in mind the requirements for Xbackwards compatability -- this will frequently disallow what would Xotherwise be desirable syntax. X XNotation: I4GL refers to both Informix-4GL and Informix-4GL RDS. R4GL Xmeans I4GL-RDS, and C4GL means (compiled) I4GL when a distinction is Xnecessary. ID means Informix-4GL Interactive Debuggger, but no suggestions Xhave been received concerning that. ISQL means Informix-SQL but does not Xfigure prominently in the discussion. The database engines are OnLine and XSE; all of the features discussed here are applicable regardless of which Xengine the application uses. A 4GL facility is a high-level, possibly Xnon-procedural, construct which is typically to do with screen or database Xhandling. A 3GL facility is normally analogous to the features found in an Xordinary programming language such as C, C++ or Ada. CDI refers to the Xnewsgroup comp.databases.informix and the mailing list. X XUnderlying assumption: it is assumed that any program which compiles and Xruns correctly under C4GL should also compile and run correctly under R4GL, Xand vice versa. X XWHAT SHOULD YOU DO WITH THIS MESSAGE? X X1: Save it. X X2: Read it. Contemplate it. Contemplate your navel. Let me know if you X find your navel helpful. X X3: If you disagree with one (or more) of the ideas in this document, send X me a message with clear identification of the idea to which you are X referring (they all have a number -- see below), and why you think it is X all wrong, and what should be done to fix it. X X4: If your favourite idea is languishing in Class D or E or F, then you X probably haven't made it sufficiently clear what you are asking for. X Please send revised details to me. X X5: If you think something is missing, put together a proposal along the X same lines as this document, and send it to me. X X6: Since it is intended to be a discussion document, please also send your X idea to the comp.databases.informix for general consideration. I will X be pleased to receive non-broadcast mail, but this is for general X discussion and as more people consider them, the ideas gradually get X improved. In this case, the more minds applied, the better. X X---------------------------------------------------------------------- X XBRIEF STATEMENT OF IMPROVEMENTS X XA quick count shows about 120 requests. These have been grouped into Xseveral subsets, and each request has been given a number of the form XI4GL-axx where "a" is a letter indicating which subset the request has been Xgrouped with, and "xx" is a 2-digit number to identify the request in the Xsubset. The number of a request does not reflect its status in any way. X X Engines I4GL-Exx X Tools I4GL-Txx X 3GL I4GL-Lxx X SQL I4GL-Sxx X Reports I4GL-Rxx X Menus I4GL-Mxx X PROMPT I4GL-Pxx X Forms I4GL-Fxx X Windows I4GL-Wxx X INPUT I4GL-Ixx X DISPLAY ARRAY I4GL-Dxx X INPUT ARRAY I4GL-Axx X CONSTRUCT I4GL-Cxx X Picklists I4GL-Kxx X XENGINES X XThese requests are really requests for changes in the database engines X(OnLine and SE), rather than the tools (I4GL, ISQL), even though they arise Xout this I4GL discussion. The list of requests here could be substantially Xextended if it were not that the focus of the discussion is I4GL. X XI4GL-E01: Control who can create databases in an OnLine system. XI4GL-E02: support for multiple databases with the same name under Online. XI4GL-E03: Name of SET EXPLAIN log file set by environment variable. XI4GL-E04: Ability to discover name of current database. XI4GL-E05: Addition of comments columns to certain system tables. XI4GL-E06: Systables should record time when table was last modified. X XTOOLS X XThese requests relate to the tools provided with I4GL. It does not address Xthe tools which should be provided with the engines. X XI4GL-T01: Library support for R4GL. XI4GL-T02: Upgrade UPSCOL. XI4GL-T03: Programmers environment consistent between C4GL and R4GL. XI4GL-T04: Programmers environment aware of forms and message files. XI4GL-T05: Programmers environment upgraded. X X3GL FEATURES X XThese requests relate to basic programming features which are more related Xto standard 3GL languages than the 4GL features of I4GL. X XI4GL-L01: Fix error -4518 properly. XI4GL-L02: General dynamic arrays. XI4GL-L03: Passing arrays by reference. XI4GL-L04: Remove size limits on strings passed to/from functions. XI4GL-L05: Proper user-defined types (especially record types). XI4GL-L06: Initialised data. XI4GL-L07: Record comparison. XI4GL-L08: Record assignment when records contain blobs. XI4GL-L09: Remove petty syntax restrictions. XI4GL-L10: Get key stroke (fgl_getkey). XI4GL-L11: Locale support for MONEY. XI4GL-L12: 8-bit characters. XI4GL-L13: Unicode support. XI4GL-L14: Multi-currency support. XI4GL-L15: Multiple globals files in a single module. XI4GL-L16: Conditional compilation and preprocessor support. XI4GL-L17: Access to __FILE__, __LINE__, __FUNCTION__. XI4GL-L18: Functions to get components of datetime or interval in integer. XI4GL-L19: USER, TIME, TODAY, CURRENT should be usable everywhere. XI4GL-L20: Explanation of interaction between TZ and DATETIME. XI4GL-L21: Private functions and variables. XI4GL-L22: Error return from STARTLOG if failed to open log file. XI4GL-L23: Better recording of errors in error log (especially user-id). XI4GL-L24: Bottom tested loop: REPEAT <statements> UNTIL <condition> XI4GL-L25: Loop-and-a-half: DO <stmts> WHILE <condition> <stmts> END DO XI4GL-L26: Allow use of DATE type as index of FOR loop. XI4GL-L27: Dynamic key mappings. XI4GL-L28: Cursors, windows, forms, attributes, key name variables. XI4GL-L29: User-defined symbolic constants. XI4GL-L30: Provide full object-orientated support. XI4GL-L31: Treat adjacent strings as a single string (as in ANSI C). XI4GL-L32: Allow defaults to be declared explicitly as well as implicitly. XI4GL-L33: Provide an ELSIF (or ELIF) keyword. XI4GL-L34: Format control for exponential notation decimal numbers. XI4GL-L35: Call functions by name. X XSQL FEATURES X XThese requests relate to modifications in the way SQL statements would be Xtreated. X XI4GL-S01: FOREACH c_cursor USING variablelist-1 INTO variablelist-2. XI4GL-S02: Support for recursive algorithms with cursors. XI4GL-S03: INPUT CURSOR and DISPLAY CURSOR. XI4GL-S04: CREATE TEMP TABLE LIKE. XI4GL-S05: UPDATE BY NAME. XI4GL-S06: Mismatch between compile-time typing and run-time typing. X XREPORTS X XThese requests relate to reports. There are some overlaps with some of the Xother sections. X XI4GL-R01: Dynamically configure size of report. XI4GL-R02: Handle multiple word-wrapped fields on one PRINT line of report. XI4GL-R03: Report with output to a 4GL window with scrolling capability. XI4GL-R04: Multiple concurrent invocations of a single report. XI4GL-R05: Control over trailing blank lines and trailing blanks on lines. X XMENUS X XI4GL-M01: Vertical menus. X XPROMPT X XI4GL-P01: PROMPT should include EXIT PROMPT and CONTINUE PROMPT XI4GL-P02: PROMPT should include BEFORE PROMPT and AFTER PROMPT XI4GL-P03: PROMPT should provide validation of data. XI4GL-P04: PROMPT "Password: " FOR passwd WITHOUT ECHO. X XFORMS X XThese requests relate to forms. There is considerable overlap with the Xfollowing sections. X XI4GL-F01: Built-in support for multi-screen forms. XI4GL-F02: Support for word-wrapped fields in screen arrays. XI4GL-F03: Eliminate the need for [|] markers in forms. XI4GL-F04: Vastly improved PICTURE attribute handling. XI4GL-F05: Horizontally scrolling char fields. XI4GL-F06: 2-dimensional screen arrays. XI4GL-F07: Control row-major or column-major ordering for matrixes of fields. XI4GL-F08: Dynamically hide screen fields (in forms) completely. XI4GL-F09: Alternative control of NEXT FIELD NEXT, NEXT FIELD PREVIOUS. XI4GL-F10: Specify "NEXT FIELD fieldname" outside scope of INPUT {ARRAY}. XI4GL-F11: Get size of current form. XI4GL-F12: Get name of current form. XI4GL-F13: Get size of screen array. XI4GL-F14: Get field names of screen record or screen array. XI4GL-F15: Document support for screen dumps XI4GL-F16: Ability to distinguish between a blank and a null field. XI4GL-F17: DISPLAY FORM should set display attributes of fields. X XWINDOWS X XThese requests relate to windows. There is considerable overlap with the Xother sections. X XI4GL-W01: MENU/PROMPT/COMMENT/ERROR LINE OFF in windows. XI4GL-W02: Get size of current window. XI4GL-W03: Get name of current window. XI4GL-W04: Allow for seamless joins between bordered windows on screen. X XINPUT X XThese requests relate to the INPUT statement. There is considerable Xoverlap with the DISPLAY ARRAY and INPUT ARRAY sections. X XI4GL-I01: Dynamic control of display attributes for individual fields. XI4GL-I02: Dynamic control of NOENTRY/ENTRY/REQUIRED for individual fields. XI4GL-I03: Simpler switch from INPUT to INPUT ARRAY in master/detail screens. XI4GL-I04: The help facility should appear in defined window. XI4GL-I05: Determine the current value of any OPTION and reuse the value. XI4GL-I06: Add field_changed() function X XDISPLAY ARRAY X XThese requests relate to the DISPLAY ARRAY statement. There is considerable Xoverlap with the INPUT and INPUT ARRAY sections. X XI4GL-D01: Dynamic control of display attributes for individual fields. XI4GL-D02: Positioning of first active row in program array. XI4GL-D03: Positioning of first active row in screen array. XI4GL-D04: Positioning of current active row in program array. XI4GL-D05: Positioning of current active row in screen array. XI4GL-D06: HELP clause for DISPLAY ARRAY. XI4GL-D07: BEFORE DISPLAY, AFTER DISPLAY clauses. XI4GL-D08: BEFORE ROW and AFTER ROW clauses. XI4GL-D09: Move faster than row/page at a time through an array. XI4GL-D10: Current row highlighted automatically. XI4GL-D11: Display a subset of the elements in a program array. X XINPUT ARRAY X XThese requests relate to the INPUT ARRAY statement. There is considerable Xoverlap with the INPUT and DISPLAY ARRAY sections. X XI4GL-A01: Ability to have current row highlighted automatically. XI4GL-A02: Dynamic control of display attributes for individual fields. XI4GL-A03: Dynamic control of NOENTRY/ENTRY/REQUIRED for individual fields. XI4GL-A04: Positioning of first active row in program array. XI4GL-A05: Positioning of first active row in screen array. XI4GL-A06: Positioning of current active row in program array. XI4GL-A07: Positioning of current active row in screen array. XI4GL-A08: Enable/disable INSERT and DELETE key. XI4GL-A09: Dynamic control of effective size of program array. XI4GL-A10: Program-controlled INSERT and DELETE operations. XI4GL-A11: Program override of user's INSERT and DELETE operations. XI4GL-A12: Move faster than row/page at a time through an array. XI4GL-A13: Display a subset of the elements in a program array. XI4GL-A14: Duplicate suppression during input. XI4GL-A15: Allow cursor out of all-blank row. XI4GL-A16: Better control of rows changed during INPUT ARRAY WITHOUT DEFAULTS. X XCONSTRUCT X XI4GL-C01: Re-instate AUTONEXT in CONSTRUCT. XI4GL-C02: Fix get_fldbuf so it returns and restores all of a condition. X XPICKLISTS X XPicklists are related to, but different from, DISPLAY ARRAY statements. X XI4GL-K01: Provide picklists. X X---------------------------------------------------------------------- X XEquivalent features: X XSome of the requests are equivalent to each other. They are either Xdirectly equivalent because they are verbatim copies of a request which Xalso appears in another context, or they are indirectly equivalent because Xwhen one of the group is implemented, the other should be implemented at Xthe same time. This list is not necessarily complete. X XSet 1: I4GL-A02 I4GL-D01 I4GL-I01 XSet 2: I4GL-A03 I4GL-I02 XSet 3: I4GL-A04 I4GL-A05 I4GL-A06 I4GL-A07 I4GL-D02 I4GL-D03 I4GL-D04 I4GL-D05 XSet 4: I4GL-A12 I4GL-D09 XSet 5: I4GL-A13 I4GL-D11 XSet 6: I4GL-L02 I4GL-L03 XSet 7: I4GL-L11 I4GL-L14 XSet 8: I4GL-L12 I4GL-L13 X XPlausibility ratings: X XClass A: I4GL-A04 I4GL-A05 I4GL-A06 I4GL-A07 I4GL-A08 I4GL-A09 I4GL-A10 X I4GL-A11 I4GL-A13 I4GL-C02 I4GL-D02 I4GL-D03 I4GL-D04 I4GL-D05 I4GL-D10 X I4GL-D11 I4GL-E03 I4GL-F02 I4GL-F13 I4GL-F15 I4GL-L01 I4GL-L03 X I4GL-L06 I4GL-L09 I4GL-L10 I4GL-L15 I4GL-L19 I4GL-L20 I4GL-L26 X I4GL-L29 I4GL-L31 I4GL-L33 I4GL-P01 I4GL-P02 I4GL-P03 I4GL-P04 X I4GL-R01 I4GL-R02 I4GL-S01 I4GL-T03 I4GL-T04 I4GL-W01 XClass B: I4GL-A01 I4GL-A02 I4GL-A03 I4GL-D01 I4GL-D06 I4GL-D07 I4GL-D08 X I4GL-E04 I4GL-E05 I4GL-E06 I4GL-F04 I4GL-F09 I4GL-F10 I4GL-I01 X I4GL-I02 I4GL-K01 I4GL-L04 I4GL-L05 I4GL-L07 I4GL-L18 I4GL-L21 X I4GL-L22 I4GL-L23 I4GL-L24 I4GL-L30 I4GL-L32 I4GL-S02 I4GL-T01 X I4GL-T02 I4GL-W02 I4GL-W03 XClass C: I4GL-A12 I4GL-A15 I4GL-E01 I4GL-F01 I4GL-F05 I4GL-F07 I4GL-F08 X I4GL-F17 I4GL-I05 I4GL-I06 I4GL-L02 I4GL-L11 I4GL-L12 I4GL-L13 X I4GL-L14 I4GL-L16 I4GL-L17 I4GL-L25 I4GL-L27 I4GL-L28 I4GL-L34 X I4GL-R05 I4GL-T04 I4GL-W04 XClass D: I4GL-A14 I4GL-E02 I4GL-F03 I4GL-I04 I4GL-L08 I4GL-R03 I4GL-L35 XClass E: I4GL-D09 I4GL-F11 I4GL-F12 I4GL-F14 I4GL-M01 I4GL-S03 I4GL-S04 X I4GL-S05 I4GL-S04 XClass F: I4GL-A16 I4GL-C01 I4GL-F06 I4GL-I03 I4GL-R04 I4GL-S06 I4GL-F16 X X---------------------------------------------------------------------- X X ************************* X ** DETAILED DISCUSSION ** X ************************* X XENGINES X XI4GL-E01: X At the moment, anybody can create their own database in an OnLine X system, and can use arbitrary amounts of disk space for their tables by X specifying suitable initial extent sizes. Informix OnLine system X administrators should be able to control who can create databases in an X OnLine system. This is so they can ensure that that valuable disk X resources are not consumed. The control should extend to allowing an X individual to create tables in some DBspaces. Similar control should X be available for blobspaces (and optical partitions). The default X control should be as lax as it is now to allow for backwards X compatability. X X I4GL-E01 was rated in class C because although it is quite important, X it is not easy to see how the permissions would be properly controlled. X XI4GL-E02: X With SE, it is possible to have multiple databases all with the same X name but located in different directories. The same suite of programs X can be used to manipulate any of the databases simply by setting the X DBPATH environment variable. This setup is particularly common for X accounting systems where there are multiple legal companies run by one X parent company on one computer system. With OnLine, you can only have X one database with a given name in any one instance of OnLine. The X direct analogue of the SE scenario would be to have multiple OnLine X systems, each with a different tbconfig file. However, the resources X required for this are very large. It is ludicrous to think of using X ths technique for a 20-company machine! The only available alternative X is to modify the programs so that they take a database name either as a X parameter or from an environment variable and open the correct X database. This leaves a minor problem when compiling I4GL programs -- X a database is needed for the non-procedural database statement. Unless X the MAIN program is compiled in a file without a non-procedural X database statement in it, some database will be opened automatically on X startup and then closed before the correct database is opened after X all. X X I4GL-E02 was rated in class D because there is neither an easy solution X nor any urgent need to implement one. X XI4GL-E03: X At the moment, the output from SET EXPLAIN ON goes into a file with a X fixed name "sqexplain.out". If several users are debugging in the same X directory at once, the traces of their programs gets interleaved, X making it unnecessarily difficult to track what was happening. The X simplest improvement would be to allow the name of the SET EXPLAIN log X file to be set by an environment variable. It would be helpful if it X each entry was given a timestamp and a user name as well. X X I4GL-E03 was rated in class A because there is an easy solution and it X is needed. X XI4GL-E04: X It is not possible to determine the name of the current database by X making enquiries from within the program. With OnLine, it would not be X very difficult to provide a dummy entry in Systables which defined the X name of the database (like the version entry and the MODE ANSI entry). X This would be more difficult to do satisfactorily in SE because the X database can be renamed by simply moving the database directory. If X the operating system supports symbolic links, then the database can X actually be known by multiple different names simultaneously! However, X when the database was created, the name could be entered, and if a X RENAME DATABASE command was provided (the subject of a completely X separate FR), it would update the entry in Systables. X X I4GL-E04 was rated in class B because there is a fairly easy solution X though it involves another minor modification to the semantics of the X system catalogue. X XI4GL-E05: X It should be possible to record in the system catalogue the reason why X a table, column or index exists by adding a comment. DB2 supports this X via the COMMENT command. Informix could support it by adding a X comments column to certain system tables. The minimum set includes X Systables, Syscolumns, Sysindexes. You can argue whether Systriggers X and Sysconstraints need extra comments, since the decoded constraint X and trigger is usually reasonably clear. Sysprocedures already has a X sort of comment facility, but it should be made uniform with the X facilities for the other comments. Maybe it should be handled by a X Syscomments table which has a serial column as primary key, and the X other tables can store a cross-reference to an entry in the Syscomments X table. This has the merit of avoiding VARCHAR fields in the main X system catalogue, and also of avoiding a frequently unused, long CHAR X column in the catalogues. X X I4GL-E05 was rated in class B because there is a fairly easy solution X though it involves another minor modification to the semantics of the X system catalogue. X XI4GL-E06: X The column Systables.Created records the date when the table was X created or last altered. It would be helpful if Systables recorded the X time when it was changed as well as the date, so that it would be X possible to tell whether software needs to be recompiled because of a X change in a table since it was last compiled. If the software was last X compiled on the same day as the table was modified, it is not possible X to tell whether it was compiled before or after the change. The X timestamp could either be a DATETIME (self-consistent) or a Unix X timestamp (an INTEGER quantity) and more readily comparable with the X timestamps on a file. X X I4GL-E06 was reated in class B because it would be easy to do and X useful. X X---------------------------------------------------------------------- X XTOOLS X XI4GL-T01: X C4GL can make use of the libraries provided by the C compiling system X on the machine where development takes place. Libraries are X fundamental to good working practices on a large project (or on X multiple projects) as they allow code to be reused. R4GL code cannot X be placed in libraries because there is no linker facility which X understands libraries. This should be provided. X X There are several design considerations to take into account. For X example, should the library containing the R4GL object files have the X same structure as the system archive or should it be proprietary to X Informix? If it uses the system archive command, then all the X different archive formats have to be understood (and there are at least X 4 possible formats to consider). However, if the system archive format X is used, the system MAKE program could then also be used to maintain X the archives since it would understand the archive format too. If a X custom Informix format was provided, then Informix would also need to X provide a version of MAKE which would understand both the Informix X archive format and the system archive format. So, it seems clear that X the system archive format should be used. X X Using the system archive means the Infomrix will have to write a X program analogous to RANLIB. RANLIB periodically made its way into and X out of AT&T Unix, and is still a feature of BSD Unix. RANLIB adds an X archive file to the front of the archive which identifies the symbols X defined in the archive, and the offset within the archive of the X archive file header for the object file defining the symbol. It is X reasonable to require that any given archive either contains R4GL X object files or C4GL (and C, ...) object files and not a mixture of the X two. If this is not done, then both RANLIB and the Informix equivalent X program will try to place their symbol file at the start of the X archive, and each of them would break down after the other had been run X (unless some very careful design was used to ensure that this was not a X problem!). Additionally, both programs would have to be run on the X archive within a short enough space of time for both of them to X consider that their symbol tables were fully up to date. That probably X means during the same second, which is not feasible in general. X X The symbol file is used by the linker to determine what files are to be X selected from the library. The linker program should emulate the X standard C loader command, taking options to indicate the output file, X the object files to be linked unconditionally, and the libraries to be X searched for other undefined symbols. Note that, unlike the loader for X a C program, the R4GL linker cannot complain about undefined symbols (by X default) because many of them are defined by the interpreter itself, X including some user-defined symbols (and such gems as STATUS, SQLCA). X X I4GL-T01 was rated in class B because the need for such support is X urgent, and because the technology required to support it is not X singularly complex. X XI4GL-T02: X Version 4.00 of UPSCOL displayed tables with the owner of the table as X a prefix if you did not own the table. Version 4.10 of UPSCOL ignores X the owner of the table. However, since you can have two tables with X the same name but different owners in a mode ANSI database, this is not X a complete solution to the problem. Actually, UPSCOL 4.10.UC3 doesn't X work on a mode ANSI database at all -- it can't even create the X validation and attribute tables. UPSCOL needs serious attention, X because it is potentially a very useful program, but it remains X seriously flawed. It would also be nice if you could use it to add X comments about the purpose of tables and columns in the absence of X I4GL-E05. X X I4GL-T02 was rated in class B because the basic technology already X exists but is not really usable because UPSCOL was never properly X upgraded from version 1.10.03X. X XI4GL-T03: X With version 4.00 of R4GL, somebody made an important improvement to X the R4GL programmers environment so that you could specify the globals X files which were used by a program. This is very necessary if you are X going to use the facilities of Syspgm4gl, but the addition was only X half thought about. You cannot specify that a given source file uses a X particular globals file. And the same change was not made to the C4GL X programmers environment. This means that you cannot use Syspgm4gl to X compile the same program under both R4GL and C4GL and get working X executables out of both programs! And you can't create two separate X program entries with the program in the same directory because the X extension for the executable is not stored in the database. Ugghh! X X I4GL-T03 was rated in class A because the incompatability between the X two versions is unforgivable. X XI4GL-T04: X Neither programmers environment is aware of forms or message files, but X a complete program does depend on having both of those available. The X programmers environment should be upgraded to support these. While the X Syspgm4gl system is being upgraded, it should support version control X software (preferably both SCCS and RCS). X X I4GL-T04 was rated in class A because the existing facility is fine as X far as it goes, but it doesn't really help on big projects. X XI4GL-T05: X Since there can only be one Syspgm4gl database in an OnLine system, X neither programmers environment really supports working on multiple X projects. For example, the pathnames typically have to be hard-coded X and absolute; this makes creating a private version of the program X under a different directory into a major exercise as there is no X facility for duplicating a program. Likewise, moving a project around X is a pain too, but in reality, people develop software on MachineA at X their office and install the software on MachineB at the customers X site, and the location of the two sets of software is seldom in the X same place, so if the customer is going to take over the maintenance of X the software, all sorts of changes have to be made. To fix this X requires a major analysis of the database and the way it is likely to X be used. X X There are other problems with the database too. For example, there is X no table which identifies programs. Programs are identified by a X composite key consisting of the program name (10 bytes) and the program X path (40 bytes); this 50-byte key is used throughout the rest of the X database. A program table should be created which has a serial column X that is used in the other tables to identify the program. X X Maybe it would be better if instead of upgrading the development X environment (beyond making the C4GL and R4GL versions compatible), the X effort was spent on providing sets of rules for use with MAKE and X guidelines on how to use that instead. X X I4GL-T05 was rated in class C because the existing facility is fine as X far as it goes, but it doesn't really help with multiple projects when X there can only be one program database in a given OnLine system. X X---------------------------------------------------------------------- X X3GL X XI4GL-L01: X This is one of a couple of bug fixes which are called out explicitly. X Error -4518 is "Out of temporary string space" or words to that effect. X This happens because (a) when strings are returned from a function, X they are limited to no more than 512 bytes each, (b) each returned X string is copied into the temporary string space, (c) the temporary X string space is not all released after every function returns, and (d) X the system only allows for up to 10 chunks of memory, each no more than X 512 bytes long. This makes it fairly easy to write code which runs out X of temporary string space, especially in 4.10 R4GL (which broke working X 4.00 C4GL code!). This problem must be completely resolved by the next X release of I4GL. X X I4GL-L01 was rated in class A because it is a bug which breaks previously X working code. X XI4GL-L02: X There are languages (though neither C nor C++ are among them) which X allow arrays to be sized dynamically. That means, in particular, that X you can call a function with an integer argument N, and the value of N X will be used to declare local arrays of dimension N. There are a few X occasions when this would be useful in I4GL. X X There are other languages (including C and C++) which allow you to X allocate new arrays dynamically. To supply this facility in I4GL would X require some modifications to the language to supply some support for X pointer types as it is very difficult to supply the functionality X without using pointers. X X I4GL-L02 was rated in class C because it would be very useful but does X not fit easily into the current model of I4GL. X XI4GL-L03: X Arrays should be made into first class variables; you should be able to X pass (by reference) arrays into functions. Not being able to do this X makes it very difficult to write some sorts of code. In particular, X library functions cannot work on user supplied arrays, so things like X generic sorting routines (even for one type of record) simply cannot be X written in I4GL. X X One of the biggest manually produced source files (as opposed produced X by a code generator) I've seen was a single file because it operated on X a single array. Because of the limitations on the number of globals X files in a single module (I4GL-L15), another globals file could not be X used to hide the information properly. So a single file was the only X solution, but it was ugly! X X I4GL-L03 was rated in class A because it would make a lot of programs very X much easier to code. X XI4GL-L04: X At the moment, there is a limit of 512 bytes on the size of the string X which can be passed into a function, or returned from a function. X Strictly, this is partly a documentation issue, because you can X actually pass an arbitrarily large character string into a function. X But, as discussed under I4GL-L01, you cannot pass arbitrarily large X strings back from a function. This is a considerable nuisance when the X program needs a function to build an arbitrarily complex query and return X the query string back again, and in other circumstances too. This problem X could be alleviated by passing a large character string into a function by X reference (so the called function could write directly on the space where X the calling function wants the answer, instead of having to copy it on to X the temporary string space stack and then copying it back off again). X X I4GL-L04 was rated in class B because it would make a lot of programs a X lot easier to write. X XI4GL-L05: X I4GL does not provide support for renaming types. This means that if X you ever have to build a record out spare parts, and you need both an X array of this type and a few single record variables, you have to be X very careful about how you edit either the declaration of the array or X the declaration of the singel variables. The only reliable way of X defining a record of the required type is to define a special view in the X database which contains no data but which has the correct structure X type, and then to define variables as shown below. This is doubly true X if the definitions are actually in two separate source files. X X DEFINE X arr1 ARRAY[30] OF RECORD LIKE Special_view.*, X var1 LIKE Special_view.*, X var2 LIKE Special_view.* X X I4GL should provide a full scale type defining mechanism. There are X enough languages to choose from when you do this (Pascal, C, Algol, X Ada, C++, ...) and it would certainly be nice to have some of the X object-orientated features of C++ added at the same time (I4GL-L30). X X I4GL-L05 was rated in class B because it would make a lot of programs X very much easier to write. X XI4GL-L06: X Most languages (including, for pity's sake, BASIC) manage to provide a X way of initializing variables without writing out screens full of X assignment statements which take up space in both the source and the X object files, not to mention in the main memory of the machine, and X which take time to execute when a better defined language would provide X the initialisation essentially for free. Both local variables and X external variables should be initializable. X X I4GL-L06 was rated in class A because it would make a lot of programs X so much easier to write and maintain. X XI4GL-L07: X Record assignments are supported (though see I4GL-L08); it would be X useful to be able to compare two records. Doing the job properly X manually is a pain, especially if you have to worry about nulls. Some X language support for this is long overdue. X X I4GL-L07 was rated in class B because it is relatively hard to do X properly, even though it would make many programs easier to write and X maintain. X XI4GL-L08: X There are a variety of restrictions on what can be done with records X containing blobs. One of the most irritating is that you cannot do X record assignment when a record contains blobs. To get around this, it X is customary to make the record holding blobs as small as possible so X that when an assignment is made, the manual copying of the other X components is either unnecessary or as limited as possible. It is not X clear what the correct semantics of copying blobs should be, especially X if the blobs are large, which is one of the main reasons why it is not X supported. X X I4GL-L08 was rated in class D because it is hard to define a sensible X set of semantics for the operator. X XI4GL-L09: X I4GL should allow any expression to be used anywhere. In C, you don't X have to worry about whether you can use the result of a function call X added to a subscripted variable as the index into an array -- of course X you can. Unfortunately, there are all sorts of petty restrictions on X what sorts of expression can be used where in I4GL. You cannot use a X subscripted variable as a subscript: array[pointer_array[i]] is not X allowed. You cannot use the result of a function call in a DISPLAY X statement: DISPLAY function_call() is not allowed. There is inadequate X distinction between the comma as a string concatenation operator and as X a list separator. (It should be better when I4GL supports the "||" X string concatenation operator which is part of Version 5.00 SQL.) X There are places where you can use a literal but not a variable; one of X the irritating ones is that the numbers indicating that the substring X to be returned in a SELECT statement cannot be variables (not even "?" X place markers) -- they have to be written directly into the text of the X SQL statement. X X I4GL-L09 was rated in class A because it is such a nuisance having to X remember where which particular forms of expression are permissible. X XI4GL-L10: X I4GL version 4.10 provided two useful functions, fgl_lastkey() and X fgl_keyval() which help a lot with identifying what on earth is going X on. However, when one wants to provide a vertical menu (amongst other X occasions), it would be very helpful to be able to get a single X character at a time. However, PROMPT FOR CHAR is no help because it X moves the cursor around the screen unpleasantly. Using INPUT on a X single character AUTONEXT field is no help because fgl_lastkey() X returns the value 2018 which stands for AUTONEXT. The existing DISPLAY X ARRAY is not flexible enough to do the required job (and probably X shouldn't be made to do it, though the upgraded version described below X would make life a lot easier than the current version does). So, I4GL X should provide a bolt-hole function fgl_getkey() which reads a X character and whose result can be interpreted using fgl_keyval(). X X I4GL-L10 was rated in class A because it is trivial to provide it and X would make life so much easier for writing programs. X XI4GL-L11: X ANSI C defines routines setlocale() and localeconv() which provide X tremendously detailed information about how money values (in X particular) should be formatted for a particular locale. There is not, X regrettably, any standard routine which formats a numeric quantity X using those rules, though there is a de facto standard in "The Standard X C Library" by P J Plauger. However, I4GL is used in many international X locations, and should should be able to provide a mechanism for X handling formatting according to locale. See also I4GL-L14. X X I4GL-L11 was rated in class C because it can be provided by diligent I4GL X programmers, but it takes considerable effort. X XI4GL-L12: X One request in CDI was specifically for 8-bit character support, X especially in ISQL forms, which has been translated to I4GL for the X purposes of this exercise, though it is desirable that ISQL forms (and X reports) should give the same support. This is presumably to enable X 8-bit characters to be entered at a terminal and displayed, and stored X in the database and retrieved, and so on. See also I4GL-L13. X X I4GL-L12 was rated in class C because it is something which is likely X to happen anyway as the engines develop support for national languages X and so on. It is possible that I4GL-L12 should really be rated X in class A. X XI4GL-L13: X The new 16-bit Unicode character set (and possibly the 32-bit character X set, whatever it is called) should ideally be supported. However, it X is far from clear what this really means, so some research is required X before this can be implemented. X X I4GL-L13 was rated in class C because it is not clear what it entails. X Maybe it should really be rated in class D or E. X XI4GL-L14: X I4GL provides the environment variable DBMONEY, but there is no provision X in the language for handling multiple currencies. Maybe the best way of X handling this is via the locale mechanism of I4GL-L11, or maybe some other X mechanism is appropriate. Some sort of support is necessary. X X I4GL-L14 was rated in class C because it is something which can be coded X around, but it would be better if there was some proper support for it. X XI4GL-L15: X At the moment, any source file can only contain one statement of the X form: GLOBALS "file.4gl". This always was a remarkably silly X restriction as there is next door to no cost in allowing zero or more X such statements compared with the cost of allowing zero or one such X statements. The processing for the second and subsequent such X statements is identical to the processing for the first. And being X able to have multiple globals files would allow far better control over X the scope of globals, since each globals file would only define closely X related variables, and each source file which needed to use those X variables would only know about the variables it really needed to know X about. In general, it would vastly improve the quality of the software X which could be produced. (See also I4GL-L16, I4GL-L17.) X X I4GL-L15 was rated in class A because it is trivial to implement and X would make an improvement out of all proportion to the difficulty of X implementation. X XI4GL-L16: X I4GL sorely lacks any facility analogous to the C preprocessor. The X language should provide support for including header files which X contain definitions of types (I4GL-L05) and constants (I4GL-L29), or X conditional compilation. X X The C preprocessor is not really appropriate for use because its X comment conventions are different from I4GL (which already has 3 X comment conventions, so a fourth, C compatible one is not a good idea), X and because it use an I4GL comment convention to indicate the X statements of interest to itself. Additionally, I4GL does not provide X support for identifying the line number in the original source file, so X the error message relating to line 3254 in the preprocessed source may X easily actually refer to line 456 in the unpreprocessed file X (I4GL-L17). X X An equivalent phase should be provided by the I4GL compilers, with X similar facilities. X X I4GL-L16 was rated in class C because it would take a certain amount of X implementation, but it would be very helpful. X XI4GL-L17: X The WHENEVER ERROR CALL function facility is not very helpful because X it is not possible to pass any information to the function, and that X includes any indication of where the error occurred. Admittedly, the X error is logged if you have an error log open (but see I4GL-L22, X I4GL-L23), and it is difficult to write the error handling code so it X reacts sensibly to the location, but then calling an error logging X function is not all that helpful because you probably have to set some X module or global variable to indicate that an error occurred, and then X check as if you had used WHENEVER ERROR CONTINUE, and since you can't X tell where the error occurred, you cannot log you error messages any X better. It would be helpful if a function called by WHENEVER ERROR X CALL was at least given the line number and file name (corresponding to X __LINE__ and __FILE__ in the C preprocessor) of the statement which X triggered the error, and the function name (__FUNCTION__?) would often X be helpful (if only because function names change less often than line X numbers in files, so different reactions could occur depending on which X function triggered the error). This could be done without change of X syntax since the existing error handling functions would simply ignore X the arguments which were provided, but new functions could use them X immediately. X X I4GL-L17 was rated in class C because it should be implemented as part X of implementing I4GL-L16, and I4GL-L16 was rated as class C. X XI4GL-L18: X There is not at present any mechanism for determining the hours part of X an interval. To do so, you have to convert the interval to just the X hours part, format that into a string, and convert the string into an X integer, which is a lot of unnecessary work. Similar comments apply to X all the other components of both interval and datetime values. With X intervals, it would also be helpful on occasion if there was a way of X converting an interval of type DAY TO FRACTION into the total number of X seconds, minutes, hours or days, where the result would be a decimal X value with an appropriate precision and scale. It is not clear whether X corresponding functions can be defined sensibly for datetime values. X These functions should also be made available to the ESQL languages. X X I4GL-L18 was rated in class B because it should not be very hard to do X and would improve the usabilty of datetime and interval types. X XI4GL-L19: X Early versions of I4GL allowed magic words like USER, TIME, TODAY and X CURRENT to be used in certain restricted contexts; TIME only in a X report, USER in a SQL statement, TODAY and CURRENT in most places. X Both USER and TIME should be usable anywhere. The other word in this X general category is SITENAME or DBSITENAME (synonyms); it is not clear X whether this should be allowed outside an SQL statement, but there is X no obvious reason why not, and it would provide uniform behaviour. X X I4GL-L19 was rated in class A because it should be easy to fix. X XI4GL-L20: X There is some interaction between the environment variable TZ and the X DATETIME value, because the US developers will not have develpoed X DATETIME working on GMT and here in GMT territory, the value of CURRENT X is normally correct. In 4.00, the interaction was not simple (it X didn't seem to change when you changed TZ), but 5.00 seems to use TZ. X This should be documented. X X I4GL-L20 was rated in class A because it is a simple change to the X documentation. X XI4GL-L21: X C provides static functions so that a file can control its external X interface. I4GL does not, but it should. While on the subject of X private objects, I4GL could also usefully provide the equivalent of X staic variables in C functions. X X I4GL-L21 was rated in class B because it should be easy to implement. X XI4GL-L22: X The function STARTLOG does not give any indication that it failed to X open the error log file. This means that when you go to use the log X file, your program stops out of control with an error about cannot X write to log file. To fix this, an alternative function is required X with a revised interface since trying to return a value where none is X expected will cause trouble in R4GL even if C4GL will survive. (See X also I4GL-L23.) X X I4GL-L22 was rated in class B because it is easy to fix. X XI4GL-L23: X The standard logged error message looks like: X X Date: 21/01/1993 Time: 15:57:41 X Program error at "rrr.4gl", line number 7. X FORMS statement error number -1110. X Form file not found X X It would frequently be helpful if at the very least the user ID and X possibly the terminal id and/or process id were also recorded, because X it is usually the case that there are multiple users who are running X the program at the same time, and it can be difficult to determine X which user it was that had a particular problem. Other information X which would also be useful is the value of ARG_VAL(0), because X sometimes different programs use a common log file. A revised format X could look like the example below. The date has a timezone appended so X that if a user in California is using a database in the UK by remotely X logging into a machine in the UK, their local time is recorded (if they X have bothered to set TZ). It avoids the problem of the log file SHAR_EOF echo "End of part 1" echo "File i4glmods is continued in part 2" echo "2" > s2_seq_.tmp exit 0