If ts is a dseries object, then

ts{'Variable_1','Variable_2'} = [];

or

ts{'Variable_@1,2@} = [];

will remove Variable_1 and Variable_2 from ts (if they exist).

Fixed bug related to the initialization of the dseries and dates object (test that the input argument is a string).

Allows the following syntax to extract variables from a
dseries object ts:

us = ts{2:end};

If ts is a dates object with 100 observations (ts.nobs=100), the
following syntax for selecting the last 90 observations *is not legal*:

ts(11:end)

ts has to be indexed instead with dates objects, as in the following example:

ts(d1:d2)

where d1 and d2 are dates objects.  To select the 90 last observations,
if the last date is unknown, the following syntax can be used:

ts(ts.dates(11:end))

Allow variables extracted from a dseries object in the dynamic or static expression following the do keyword.

Use new leadlagtable's columns introduced in d0708fcb (usefull for sparse dynamic models).

Another speed improvement for @dates/colon method (vectorized code, removed loop around add_periods_to_date routine).

Improve speed of @dates/colon method (added specialized version of add_periods_to_array_of_dates -> add_periods_to_date).

Save empty dates and dseries objects in memory (base workspace) instead of disk (faster). Added an initialization mode for dates and dseries constructor. The following commands:

dates('initialization');
dseries('initialization');

create variables emptydatesobject and emptydseriesobject in base workspace.

Exclude from versioning the mat files generated for the initialization of dseries and dates objects.

Save empty dates and dseries objects on disk (in dynareroot), to reduce the time needed to instantiate new objects.

main loop.

canot depend on its current level or its future level).

dseries objects appearing in the recursive expression are compatible
with the range of the loop.

dseries objects involved in the recursive expression.