On Mon, 22 Dec 2014, Ignacio Diaz-Emparanza wrote:
This is only to note that I have just tried to run a script that
was correctly working on 2012:10 and with gretl from the current
CVS this was not working now.
The three problems I found were in the same command:
model$i-$j <- hsk $i 0 xepl --quiet
1) The option '--quiet' is not valid now for hsk.
Hmm, I'm not seeing a problem with that. For example,
hsk 1 0 2 3 4 --quiet
works fine for me.
2) The name of the model cannot contain a '-'
OK, I see that's an effect of the new tokenizer. However, I think
it's reasonable that if a model name is not a valid identifier it
should be quoted. So
"model-foo" <- ols ...
works alright -- as does the following, which is closer to what you
"model-$i" <- hsk $i 0 6 --quiet
3) (This is not a problem, is an advantage of the new situation)
The new command interpreter detected a problem with nested
loops that were using the same i index (The inner i loop was
already finished when executing the j loop). In 2012 this was
being processed and giving the correct results.