Re: [Oorexx-devel] open() option APPEND
On 2021-02-10 15:21:18 CVBruce wrote: > Eric, > > To your point #2. In looking at other implementations of Rexx, I found > that some implementations note that if you move the write pointer to > someplace in the existing file, and then write, the file will be truncated > to that written data. This led me to the realization that the write > pointer is also the “working” end of file pointer. If you move the write > pointer to the middle of the file what will the result be, especially if > the data written is greater than the original data in the position, wiping > out the record separator (CRLF) in the process. > > I would lean toward not being able to move the write pointer, and use > sequential file update processes that are old and well defined. Ah. That's right, I forgot about that truncation issue. > > I know that update in place is a common technique on the mainframe, but > mainframes also support record structures, and record i/o, Rexx for the > most part doesn’t. There is not a data structure that is tied to the file > in away that prohibits writing records that don’t agree with the structure > of the file. On the mainframe (where Rexx was first implemented), update-in-place only worked for fixed-length record files; until the Byte File System in VM and Unix System Services in MVS were implemented, variable-length record files had a record-length prefix on each record; there was no concept of "stream-of-bytes with imbedded linend characters". > > To your point #1, I think replace is more common than default of append. > In most programming environments I’ve used you must explicitly specify > append. > > There have been times in Rexx, where to bullet proof a program, I’ve had to > issue a close then open against a file because the program in development > never reached the file close logic from the last execution, the result > being an append rather than the desired replace. > > Bruce Leslie -- ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
Eric,
To your point #2. In looking at other implementations of Rexx, I found that
some implementations note that if you move the write pointer to someplace in
the existing file, and then write, the file will be truncated to that written
data. This led me to the realization that the write pointer is also the
“working” end of file pointer. If you move the write pointer to the middle of
the file what will the result be, especially if the data written is greater
than the original data in the position, wiping out the record separator (CRLF)
in the process.
I would lean toward not being able to move the write pointer, and use
sequential file update processes that are old and well defined.
I know that update in place is a common technique on the mainframe, but
mainframes also support record structures, and record i/o, Rexx for the most
part doesn’t. There is not a data structure that is tied to the file in away
that prohibits writing records that don’t agree with the structure of the file.
To your point #1, I think replace is more common than default of append. In
most programming environments I’ve used you must explicitly specify append.
There have been times in Rexx, where to bullet proof a program, I’ve had to
issue a close then open against a file because the program in development never
reached the file close logic from the last execution, the result being an
append rather than the desired replace.
Bruce
> On Feb 10, 2021, at 11:21 AM, Erich Steinböck
> wrote:
>
> Thanks for clarifying the timeline :-)
>
> Now back to the question:
> s~open('append'); s~lineOut('xxx', 1) ignores the line number argument for
> lineOut and unconditionally appends
> s~open; s~lineOut('xxx', 1) overwrites the first line
>
> This is pretty crazy, as
> 1) "append" should be default (or not?) for any write operation, and
> 2) if "append" prohibits overwriting existing lines, we can never overwrite
> parts of an existing file at all
>
> I'd propose to remove the rewrite restriction and make "append" default, but
> then: we also have this rexxref paragraph "The write pointer cannot be moved
> anywhere within the extent of the file as it existed when the file was
> opened."
>
> So many errors in one place?
> Thus my question: anyone remembers what the intention of all this was?
>
>
>
> ___
> Oorexx-devel mailing list
> Oorexx-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/oorexx-devel
signature.asc
Description: Message signed with OpenPGP
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
Greetings Erich, Erich Steinböck wrote: but then: we also have this rexxref paragraph "The write pointer cannot be moved anywhere within the extent of the file as it existed when the file was opened." So many errors in one place? Thus my question: anyone remembers what the intention of all this was? I believe opening a file in "append" mode is special in that it protects the file content already on disk with a sort of virtual partition. That pre-existing part you may read, just not make changes to. Anything appended to the file while the file is open in append mode is far game for write/read/remove. The program is free to move around within the file reading... just the virtual partition comes firmly into place for write/delete operations. I do not know history on this "append" mode topic. I think it is a solid functionality to maintain in the ooRexx language. Is it causing you some sort of technical issues? I am thankful, -- Michael Lueck Lueck Data Systems http://www.lueckdatasystems.com/ ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
From my collection of old manuals, the 1982 version of Rexx (3.10) recommended EXECIO or IOX (which it notes was "specially written for Rexx") for file I/O. Stream/linein/lineout, etc. aren't there. The Stream/linein/lineout,etc. are not in the 1986 VM/SP Release 5 System Product Interpreter manual either. There was a 1992 package for AIX from Endicott that includes Stream, Linein, lineout, etc. including the replace & append options Stream, Linein, Lineout, charin, charout, etc. appear to be fully documented in the VM/ESA 2.1 manual from 1993. However, unlike the 1992 AIX package "append" is not documented as an option on the stream open command. Only read/write/new/replace are listed as "intents". The summary of changes at the front of the manual indicates that the I/O functions were added to VM/ESA 2.1 in June, 1993 and existed in a previous version of Rexx. The wording in the AIX description of append is similar to the description in rexxref. It says the write pointer is positioned at the end of the stream and you cannot position the write pointer to any position that already existed when the file was opened. It doesn't say that the write location operand of charout or lineout is ignored. That sounds to me like it might have allowed you to re-write any new lines you appended but not any old, existing lines. It also doesn't say what happens if you try to re-write an old location i.e. does it append anyway? or does it return a non-zero result indicating the chars/line was not written? Chuck Berghorn oorexx-devel@lists.sourceforge.net On 2/10/21 12:56 PM, Mike Cowlishaw wrote: The Stream functions were added to Rexx in 1982 (I think) but didn't make the cut for the first VM/CMS release (SP3, 1983). They are in the first edition of The Rexx Language (1985). It took a frustratingly long time for them to be added -- no quite sure why that was. Mike -Original Message- From: J Leslie Turriff [mailto:jlturr...@mail.com] Sent: 10 February 2021 16:36 To: Open Object Rexx Developer Mailing List Subject: Re: [Oorexx-devel] open() option APPEND You may be right; I was unable to find a copy of the pertinent Rexx User's Guide from that period, though http://www.rexxinfo.org/All%20IBM%20Rexx%20manuals/ALLIBM_1.HTM purports to provide such. Leslie On 2021-02-10 04:30:20 René Jansen wrote: Hi Leslie, I am sure you mean 1984. And then there was still only EXECIO. I have to look up when VM/CMS got stream; TSO never had it until now, unless we installed a library. I will try to find out the timeline. best regards, René. ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
Thanks for clarifying the timeline :-)
Now back to the question:
s~open('append'); s~lineOut('xxx', 1) ignores the line number argument for
lineOut and unconditionally appends
s~open; s~lineOut('xxx', 1) overwrites the first line
This is pretty crazy, as
1) "append" should be default (or not?) for any write operation, and
2) if "append" prohibits overwriting existing lines, we can never overwrite
parts of an existing file at all
I'd propose to remove the rewrite restriction and make "append" default,
but then: we also have this rexxref paragraph "The write pointer cannot be
moved anywhere within the extent of the file as it existed when the file
was opened."
So many errors in one place?
Thus my question: anyone remembers what the intention of all this was?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
The Stream functions were added to Rexx in 1982 (I think) but didn't make the cut for the first VM/CMS release (SP3, 1983). They are in the first edition of The Rexx Language (1985). It took a frustratingly long time for them to be added -- no quite sure why that was. Mike > -Original Message- > From: J Leslie Turriff [mailto:jlturr...@mail.com] > Sent: 10 February 2021 16:36 > To: Open Object Rexx Developer Mailing List > Subject: Re: [Oorexx-devel] open() option APPEND > > You may be right; I was unable to find a copy of the > pertinent Rexx User's Guide from that period, though > http://www.rexxinfo.org/All%20IBM%20Rexx%20manuals/ALLIBM_1.HTM > purports to provide such. > > Leslie > > On 2021-02-10 04:30:20 René Jansen wrote: > > Hi Leslie, > > > > I am sure you mean 1984. And then there was still only > EXECIO. I have > > to look up when VM/CMS got stream; TSO never had it until > now, unless > > we installed a library. I will try to find out the timeline. > > > > best regards, > > > > René. ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
You may be right; I was unable to find a copy of the pertinent Rexx User's Guide from that period, though http://www.rexxinfo.org/All%20IBM%20Rexx%20manuals/ALLIBM_1.HTM purports to provide such. Leslie On 2021-02-10 04:30:20 René Jansen wrote: > Hi Leslie, > > I am sure you mean 1984. And then there was still only EXECIO. I have to > look up when VM/CMS got stream; TSO never had it until now, unless we > installed a library. I will try to find out the timeline. > > best regards, > > René. > > > On 10 Feb 2021, at 05:45, J Leslie Turriff wrote: > > > > On 2021-02-09 19:02:12 Michael Lueck wrote: > >>> Any ideas on how this was meant to be when it was designed? > >> > >> I never claim to be the best expert at REXX history, however... > >> > >> On Mainframe, I believe either the Stream concept came much later or > >> still is not there at all. > > > > The file I/O functions [stream(), charin(), charout(), chars(), > > linein(), lineout(), lines()] have been in the language since at least > > when I first used VM/CMS in ~1974. They could operate on CMS minidisk, > > Shared File System*, or Byte File System* files; sockets capabilities > > came later when a TCP/IP stack was added to VM. In MVS Rexx could use > > these functions to do I/O to QSAM, Open Edition or Unix System Services > > files. See https://www.vm.ibm.com/library/640pdfs/64622201.pdf Chapter 9: > > Input and Output, for example. > > > > Leslie > > > > > > -- > > > > > > ___ > > Oorexx-devel mailing list > > Oorexx-devel@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/oorexx-devel > > ___ > Oorexx-devel mailing list > Oorexx-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/oorexx-devel -- ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
Greetings René, René Jansen wrote: I am sure you mean 1984. And then there was still only EXECIO. Excellent correlation! Indeed EXECIO was the mainframe I/O alternative I could not recall the name of. I am thankful, -- Michael Lueck Lueck Data Systems http://www.lueckdatasystems.com/ ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
On Wed, Feb 10, 2021 at 5:31 AM René Jansen wrote: > Hi Leslie, > > I am sure you mean 1984. And then there was still only EXECIO. I have to > look up when VM/CMS got stream; TSO never had it until now, unless we > installed a library. > I will try to find out the timeline. > VM/CMS got it 1988, at the earliest. Rick > > best regards, > > René. > > > > On 10 Feb 2021, at 05:45, J Leslie Turriff wrote: > > > > On 2021-02-09 19:02:12 Michael Lueck wrote: > >>> Any ideas on how this was meant to be when it was designed? > >> > >> I never claim to be the best expert at REXX history, however... > >> > >> On Mainframe, I believe either the Stream concept came much later or > still > >> is not there at all. > > > > The file I/O functions [stream(), charin(), charout(), chars(), > linein(), lineout(), > > lines()] have been in the language since at least when I first used > VM/CMS in ~1974. > > They could operate on CMS minidisk, Shared File System*, or Byte File > System* files; > > sockets capabilities came later when a TCP/IP stack was added to VM. In > MVS Rexx could > > use these functions to do I/O to QSAM, Open Edition or Unix System > Services files. > > See https://www.vm.ibm.com/library/640pdfs/64622201.pdf Chapter > 9: Input and Output, for > > example. > > > > Leslie > > > > > > -- > > > > > > ___ > > Oorexx-devel mailing list > > Oorexx-devel@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/oorexx-devel > > > > ___ > Oorexx-devel mailing list > Oorexx-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/oorexx-devel > ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
Re: [Oorexx-devel] open() option APPEND
Hi Leslie, I am sure you mean 1984. And then there was still only EXECIO. I have to look up when VM/CMS got stream; TSO never had it until now, unless we installed a library. I will try to find out the timeline. best regards, René. > On 10 Feb 2021, at 05:45, J Leslie Turriff wrote: > > On 2021-02-09 19:02:12 Michael Lueck wrote: >>> Any ideas on how this was meant to be when it was designed? >> >> I never claim to be the best expert at REXX history, however... >> >> On Mainframe, I believe either the Stream concept came much later or still >> is not there at all. > > The file I/O functions [stream(), charin(), charout(), chars(), > linein(), lineout(), > lines()] have been in the language since at least when I first used VM/CMS in > ~1974. > They could operate on CMS minidisk, Shared File System*, or Byte File System* > files; > sockets capabilities came later when a TCP/IP stack was added to VM. In MVS > Rexx could > use these functions to do I/O to QSAM, Open Edition or Unix System Services > files. > See https://www.vm.ibm.com/library/640pdfs/64622201.pdf Chapter 9: > Input and Output, for > example. > > Leslie > > > -- > > > ___ > Oorexx-devel mailing list > Oorexx-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/oorexx-devel ___ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
