From: Stefan Hoffmeister <Stefan.Hoffmeister@Econos.de>

: On Thu, 21 Oct 1999 12:48:49 -0400, Michel Pelletier wrote:

...

...

From: Stefan Hoffmeister [mailto:Stefan.Hoffmeister@Econos.de]

...

...

(Re)starting ZServer takes about 18 seconds on my development NT4 system (192 MB, 540 Mhz overclocked Celeron). I somehow feel that this is an a tad bit inappropriate time lag.

Let me add "for development purposes" to clarify.

Look, I am used to blazingly fast compilers - Object Pascal (Borland Delphi) and Borland C++Builder with well-done pre-compiled headers.

19 seconds delay, combined with the "manual" action of shutting down the server + starting it up again - I perceive that as an eternity.

What do I do in these 19 seconds?

...

the ability to do a python reload could be elaborated and implimented, of course you have to source code so you can provide a patch. ;)

I'm having the same problem using Python code as an external method. The trouble is that when you change the code and rrun it, the Python interpreter has continued to run all the while (since it is running ZServer). It thinks that all your code has already been bound to the existing method and class names you defined the first time you ran the code, and it doesn't realize there have been changes. A Python reload() call, as mentioned above, can work if you have imported a module using "import spam". You could put an import spam/reload(spam) pair into the function call. But reload() doesn't work for all flavors of the import statement. In my case, some of my code can be changed on the fly, and for other parts I have to restart the server. Perhaps some Python expert can give us working examples (not guesses!) showing how to construct modules and import trees so they will always update without us having to restart the server. On the other hand, those reload() calls would take time to execute, so it should be possible to remove them for the production system easily and without ill effects. Thomas Passin