Averager class does not support removal of data

[Jari Häkkinen]

The 'weight' for Averager::add is an unsigned long and there is no remove function. Either we need to change unsigned long to signed long or add a remove function.

[Peter]

Ok, I think changing unsigned long to long is preferable. I think we can keep the member variable n_ unsigned.

We should also change AveragerPair? so it have the same interface.

[Jari Häkkinen]

I have already fixed Averager (needed it urgently). I'll check in Averager and work on AveragerPair.

[Jari Häkkinen]

changeset:1290: Averager accepts negative n for removal of data from the Averager.

[Jari Häkkinen]

changeset:1291: AveragerPair accepts negative n for removal of data.

[Jari Häkkinen]

This change breaks rules for interface changes in a stable branch, moved to next minor release.

[Jari Häkkinen]

changeset:1294 reverts 0.4-stable to state before changes related to this ticket.

[Jari Häkkinen]

changest:1295 resolves this ticket.

[Peter]

./averager_test -v results in

testing Averager
lt-averager_test: Averager.cc:54: void theplu::yat::statistics::Averager::add(double, long int): Assertion `n>-1' failed.
Aborted

[Jari Häkkinen]

Replying to peter:

./averager_test -v results in

testing Averager
lt-averager_test: Averager.cc:54: void theplu::yat::statistics::Averager::add(double, long int): Assertion `n>-1' failed.
Aborted

I cannot reproduce this error. 'make check' works fine and so do './averager_test -v'. I am on a MacOSX Leopard with Intel processors ... reading the code I realize that the assert should react ... there is a test for n==-2 that is supposed to return true! Ah, I have no debug options to my compiler. Sorry about that, I'll rewrite the test.

[Jari Häkkinen]

changeset:1297 resolves the assert() issue.