On Mon, Jun 20, 2022 at 6:53 AM Sven Schreiber <svetosch(a)gmx.net&gt; wrote: This sort of thing cannot work. The non-responsiveness is now "fixed" in git and snapshots, in that an attempt to open a session file within a loop produces an error message. Besides using the GUI proper, you can open a session file via command in the console, or at the top level of a script, but not in a loop. By design, opening a session file clears out then repopulates gretl's entire workspace, and doing so within a loop is bound to destroy the integrity of the loop itself: after the first iteration the loop-counter variable will have disappeared. Allin

On Mon, Jun 20, 2022 at 9:34 AM Sven Schreiber <svetosch(a)gmx.net&gt; wrote: One way to allow access to things inside a session file without terminating the current session might be to implement opening such a file as a bundle. Saved models would appear as bundles inside the main bundle. That approach could work OK in a loop. Allin

On Mon, Jun 20, 2022 at 11:11 AM Cottrell, Allin <cottrell(a)wfu.edu&gt; wrote: Perhaps Fred could tell us if the following is useful (it's now in git and snapshots, but not yet documented). You can open a gretl session file as a bundle, via the syntax open foo.gretl --bundle=bname where "bname" is the name of the bundle to be created. At present the only relevant content of the bundle is an array named "models", holding the models defined in the session. In what I take to be the spirit of Fred's usage, here's a trivial example. A first script simulates production of a number of session files by students. Notionally, this is in response to a directive along the lines of: "Open Ramanathan data4-1 in gretl and estimate two models via OLS. In each case the dependent variable is price. In the first case the sole regressor is sqft and in the second the two regressors are sqft and bedrms. A y-intercept should be included in both cases." <hansl> /* construct a presumed correct session file, plus a few correct or incorrect "student" session files */ # authorized version open data4-1 m1 <- ols 1 0 2 m2 <- ols 1 0 2 3 store authorized.gretl # student 1 (all correct) clear open data4-1 m1 <- ols 1 0 2 m2 <- ols 1 0 2 3 store student1.gretl # student 2 (incorrect: const missing) clear open data4-1 m1 <- ols 1 2 m2 <- ols 1 2 3 store student2.gretl # student 3 (incorrect; model 2 missing) clear open data4-1 m1 <- ols 1 0 2 store student3.gretl </hansl> A second script opens the session files in sequence and checks the "students'" results: <hansl> set verbose off clear # the presumed correct results open authorized.gretl --bundle="b0" scalar nm0 = nelem(b0.models) # number of session files to check scalar n_students = 3 loop i=1..n_students # construct name of session file to read sname = sprintf("student%d.gretl", i) # construct name of bundle to write bname = sprintf("b%d", i) # open the session file as bundle open @sname --bundle="@bname" # get the array of models from bundle bundles ma = @bname.models # and determine the number of models scalar nm = nelem(ma) string s = (nm == nm0)? "OK" : "wrong" printf "student %d, nmodels=%d, %s\n", i, nm, s # compare models in terms of number/values of coeffs scalar nmi = nm < nm0 ? nm : nm0 loop j=1..nmi scalar nc = b0.models[j].ncoeff if ma[j].ncoeff == nc matrix c0 = b0.models[j].coeff matrix ci = ma[j].coeff printf " model %d: ncoeff correct\n", j printf " max diff = %g\n", max(diff(abs(c0-ci))) else printf " model %d: ncoeff wrong\n", j endif endloop endloop </hansl> I get the following results after running the two script above: <output> student 1, nmodels=2, OK model 1: ncoeff correct max diff = 0 model 2: ncoeff correct max diff = 0 student 2, nmodels=2, OK model 1: ncoeff wrong model 2: ncoeff wrong student 3, nmodels=1, wrong model 1: ncoeff correct max diff = 0 </output> Allin