Hackers' Corner

The opinions expressed herein are the author's and should not be disseminated beyond the known universe in any form including flushing down a public drainage system

Contents List

• Headbones in DEC FORTRAN/F90
• Variable formats
• OTS$CVT_T_D function
• Headbones in Microsoft PowerStation F90
• Functions in WRITE lists
• Dynamic-length AUTOMATIC strings don't work properly via ENTRY points
• "Intentional" differences between DEC and Microsoft F90
• Experiments in process
• Moving up to FORTRAN 90
• Migration from VAXen to PCs
• Impressions of Windows 95
• Impressions of Internet Explorer vs. Netscape Navigator
• Impressions of Netscape Communicator
• NOT! the Starlink Programming Standards
• Programming style
• Processing duff dates
• Some GALAXY Secrets
• Collected ramblings about overview analysis
• The Story of Forest Plots

Headbones in DEC FORTRAN/F90

Many earlier headbones have now been smashed - well done DEC!

Variable formats

      DIMENSION KK(4)
      MM=4
      DO I=1,MM
        KK(I)=I
      ENDDO
      WRITE(6,60) (KK(I),I=1,MM),MM
60    FORMAT(I5,<MM>I5)
      END

    1    4

    1    2    3    4    4

OTS$CVT_T_D function

      REAL*8 DD
      CHARACTER*6 CH/'12.345'/
      J=OTS$CVT_T_D(CH,DD,,,,)
      WRITE(6,*) DD
      END

  46213848104.9600

  12.34500

Headbones in Microsoft PowerStation F90

This section comprises some immediate headbones on starting to use PowerStation 4. These have not yet been explored or circumscribed (in extremis! shall come back to them sometime) - they might or might not prove reproducible and/or reducible.

Functions in WRITE lists

      CHARACTER*100 CH

...
      J=MAX0(VERIFY(CH,' '),1)
      WRITE(6,*) TRIM(CH(J:))
! so far, so good
      WRITE(6,*) TRIM(CH(MAX0(VERIFY(CH,' '),1):))
! the nesting fails

Dynamic-length AUTOMATIC strings don't work properly via ENTRY points

      SUBROUTINE X(CH)
      CHARACTER*(*) CH
      CHARACTER*(LEN(CH)),AUTOMATIC :: CC
! so far, so good
      ENTRY XX(CH)
! CC often has length 0 when calling XX

Anyone fancy calling Microsoft's Help-Line to tell them about it (and pay them $150)?

"Intentional" differences between DEC and Microsoft F90

• Dynamically-allocated strings can have a dynamic length specification in DEC F90 but not in Microsoft F90 (remedy: use AUTOMATIC instead of ALLOCATABLE, although this doesn't work properly via ENTRY points (above))
• The DEC FORALL statement does not have a counterpart in Microsoft (remedy: make do with WHERE etc)

Experiments in process

Moving up to FORTRAN 90

Some of the best new features, that have immediately led to simpler/better code and increased flexibility, are:

• The TRIM and LEN_TRIM functions - it is truly amazing that LEN_TRIM was not mandatory in previous incarnations of FORTRAN, as the first thing anybody who needs to produce a legible print-out will want to know is how many characters of stuff there are in a string variable.
• Global assignments to arrays - avoid many tedious DO loops to initialise arrays, etc
• WHERE and, even better (but tragically DEC-specific), FORALL - avoid many more tedious DO loops and feel the buzz of parallel processing (or same being simulated, who knows)
• Useful new intrinsic functions for array-handling (find largest/smallest number, sum rows or columns etc)
• AUTOMATIC variables (data are stored on a stack instead of in fixed locations, presumably giving better performance and/or less wasted brainspace - some techie will know), instantly applied to all relevant variables in the program by using the /RECURSIVE switch to the F90 compiler command. Beware that when you exit from a routine the values of the local variables will be lost - but if you have never trusted them to be kept (possibly a relict of experience with old, bad compilers) this makes no difference as you will, of course, be keeping everything important in COMMON... The point of all this is that you can now declare strings and arrays dynamically as you need them without getting involved with the truly unportable horrors of LIB$GET_VM and its pals. Also, for ALGOL exiles, there is recursion at last.

Migration from VAXen to PCs

This is being accomplished using Microsoft FORTRAN PowerStation 4 running under Windows NT 3·51, the equivalent of methadone to the heroin of VMS. Having put together a library of sundry routines to replace VMS Run-Time Library calls (LIB$, STR$, SYS$ etc), the transition is proving quite painless. There is an excellent mathematical and statistical library 'IMSL' which makes it easy to replace calls to the NAG Library. The VMS logical name table is simulated (experiments to determine the better method - symbols or registry - are in process) to make filestore navigation simple. A few headbones have surfaced (above).

It would be interesting to set up a client-server architecture - for instance, all the overviews in GALAXY ('frozen' Childhood A.L.L., Breast Cancer, Colorectal Cancer, Hodgkin's Disease, Prostate Cancer etc etc) could then be accessed and graunched over the CTSU Intranet. (Experiments are in hand on the Internet for GALAXY, non-CTSU work and details will be posted later). The GALAXY Address Finder,GALAXY Fortune Cookies and the GALAXY ICD Decoder are some experimental thingies on the Intranet - sorry if the (pirate) server is not switched on [please note that you will probably need to add p80.ctsu.ox.ac.uk to the list of domains which don't go via the proxy server].

Impressions of Windows 95

Partly, it's the gratuitous new interface - not a problem per se; but so many things are 'hidden' or 'done behind your back'. If you are a control freak and actually like to know where your files are and to be informed before huge swathes of strange directories and stranger associations appear spontaneously all over the place (in case you would like to get rid of them again sometime), this could be a problem.

Yes, it has built-in toolbars; but we all have Microsoft Office's toolbar, don't we? Yes, Windows 95-aware software self-installs (rather too eagerly) without your needing to make real decisions. However, try to get rid of certain tiresome ikons from the desktop, or find that the DOS Box has yet again been collapsed to a button when the screen saver cuts in, and you begin to wonder what the advantages actually are. What about that old '386' at home? - well, you put in more RAM and drink a lot of coffee whilst waiting for things to happen.

Of course it is less crashable than the old Windows 3·1/3·11 (although far from bombproof), and quite a few new applications actually insist on something more up-to-date - but Windows NT 3·51 seems far better (traditional interface, virtually unburstable, loads of serious features if required).

'Here comes the twister': there is now Windows NT 4, with the atrocious new interface - and applications are coming out that insist on Windows 95 or NT 4, for some unfathomable reason (e.g. AltaVista Search My Computer), whereas the upgrade from NT 3·5 to 3·51 was supposed to deal with all that (e.g. the Windows 95 application Microsoft Office Professional '7' works perfectly on NT 3·51). Ho hum, plus ça change etc etc.

"It's just your age, dear"

Afterburn: the gossip is that Windows NT 4 was a rather hasty lash-up just to incorporate the Windows 95 interface, and that the 'definitive' new version will be NT 5, which is nearing completion; NT 5 will include numerous features that they didn't have time to bung into NT 4. If i were to upgrade from NT 3·51 (heaven forfend), i'd definitely wait for NT 5 - unless it is going to resemble Internet Exploder (see below).

Impressions of Internet Explorer vs. Netscape Navigator

The two nicer-than-Navigator-3 features of Explorer spotted so far are: a (large, expandable) History List that can be used to add to the Favourites List; and slightly neater-looking form controls. That's it! and in any case Navigator 4 has caught up (below). Some of the nastier features are: slower loading of documents, worse rendition of colour pictures (JPEG and GIF, at least), sluggish movement of GIF animations, gung-ho font size control, slower Java execution, pop-up message 'the script may have hung: do you wish to continue?' sometimes when exiting from Java subroutines and the complete absence of a document editing suite (no doubt that could be found somewhere (uhh - FrontPage, presumably, if you don't mind alienating all of your audience who don't have a T4+ connexion) and added on, but why bother? - possible answer: it might be possible to edit form controls ('widgets'), e.g. Command and Radio Buttons, Check Boxes, Text Boxes and Areas etc, something that Navigator's otherwise excellent editor rather cops out on - does anybody know? - rsvp: skull@wolfbane.com ). Explorer is no spoiler for Navigator, being far worse than Netscape's betas. Perhaps it would be happier with 32 Mb of R.A.M.

The rumour is that the interface of the successor to Windows 95 (and, presumably, Windows NT 4) is going to resemble Internet Explorer too. If this is the face of Microsoft's world domination plan, hooray for its probable doom.

Impressions of Netscape Communicator

Being larger and more complex, Communicator is slower than Navigator 3 - indeed, clunky on first impression, although one rapidly finds the best ways of working. There is a customisable Personal Toolbar, a better Favourites List, proper font control (in Composer), a decent History List and the 'widgets' look slightly neater; there is even a Spell Checker, for persons who have the patience to fill the Personal Dictionaries of these things with the entire contents of the Oxford Textbook of Medicine, the Times Atlas of the World etc etc. The mail package (Messenger) seems improved too. The bugs spotted in previous 'preview' versions of Composer all seem to have been fixed in this final release.

Overall verdict: a very nice one, Netscape! As a postscript, however, it should be noted that trying to edit files larger than a couple of hundred kb was evidently never envisaged. With a 100 MHz Pentium and 24 Mb R.A.M. it can easily take a couple of hours to make the smallest change to a rather larger file - much longer if Autosave is inadvertently left active - never mind the real-time TV processing, this calls for one of those multi-GHz dream machines!

NOT! the Starlink Programming Standards

Processing duff dates

Some GALAXY Secrets

The Story of Forest Plots

[Link to the CTSU GALAXY Experimental Page - please note that this server might not be switched on!]

[Return to the GALAXY Software Page]

[End of document of 22 July 1998, updated]