Wsrtjtyk

In my task I need to run the same GAMS model with different parameters values in Python. For that I created GAMS workspace, model, job and GDX database:

model = '''Variables 

        objective, {vars};

    Parameters

        {up_bounds}; {low_bounds};

    $gdxin db

    $load {load_values}

    $gdxin

    Equations  

        obj, {eqs_list};

        obj ..    objective =e= {obj_equation};

        {nodes_eq};

        {up_bounds_eq};

    Model myLP /all/;

    Solve myLP using lp minimizing objective;'''.format(load_values=' '.join(load_values), kwargs**)

ws = GamsWorkspace(working_directory='some_path')



db = ws.add_database('db')

load_values_p = [self.db.add_parameter(value, 0).add_record() for value in load_values]



job = ws.add_job_from_string(model)



for i in range(any_count):

    update_parameters_values()

    job.run(databases=self.db)

    result.append([job.out_db[key][()].level for key in res_values_names])

where:
load_values - list of variables names, which will be load from 'db.gdx'. They also initialized in Parameter section of the model;
update_parameters_values() - function which change load_values_p values (it also automatically changes in db object and db.gdx file);
res_values_names - list of result variables names.

The problem is that job generates new out_db GDX file on every iteration, but I want to rewrite existing one and just read updated result values. How to do that? Thanks)

I found solution in GAMS documentation - GamsModifier can solve my problem. Applying it to my task and code above:

model = '''Variables 

        objective, {vars};

    Scalars

        {up_bounds}; {low_bounds};

    Equations  

        obj, {eqs_list};

        obj ..    objective =e= {obj_equation};

        {nodes_eq};

        {up_bounds_eq};

    Model myLP /all/;'''.format(kwargs**)

ws = GamsWorkspace(working_directory='some_path')



cp = ws.add_checkpoint()

master_init_job = ws.add_job_from_string(model)

master_init_job.run(checkpoint=cp)

mi = cp.add_modelinstance()



load_values = [mi.sync_db.add_parameter(load_value, 0)

               for load_value in load_values]



modifiers = [GamsModifier(load_value) for load_value in load_values]

mi.instantiate('myLP using lp minimizing objective', modifiers=modifiers)



for load_value in self.load_values:

    load_value.add_record()



for i in range(any_count):

    update_parameters_values()

    mi.solve()

    result.append([mi.sync_db.get_variable(key)[()].level 

                   for key in res_values_names])

At the time of the start, the values of all variables are similar to listed in the question.
This implementation avoids the creation of additional files and works faster.