Re: Tcl bytecode

Jeff Hobbs wrote:

No, it actually uses the same bytecodes. However, TDK has to make the
bytecode portable (no BE vs. LE issues allowed), so that is part of the
extra translation that doesn't make this loader faster.

<snip>

One main reason that this hasn't been an issue is that there isn't a
huge overhead to compiling. Other languages have more compile overhead,
so it makes more sense there to precompile. Perl for example likes to
do a full syntax parse, even over code areas that won't be used. Tcl is
much lazier - it only compiles procs as it first executes them. This
distributes any compile overhead over the runtime of the application.

Some large Tcl apps are slow to start: is this because too many procs are
executed in the startup script?

Python seems to have a good compromise: when it is installed for the first
time, all the libraries are compiled and saved locally as *.pyc files
alongside the *.py originals. This way there are no portability issues.
Also, tampering with the *.pyc file is presumably no more difficult than
tampering with the original *.py.

Keith.

.