IdBuild Macro-Equation Processor for InterDymeΒΆ

IdBuild is an adaptation of the Build program, which is used for aggregate model building, to serve the construction of Interindustry Dynamic models. Like Build, it translates the .SAV files created by G7 into C++ code, builds a bank of all the scalar variables, and writes several files of C++ code for use in the simulation program. Also like Build, it requires a BUILD.CFG file that specifies the name of the output (or workspace) bank and the initially-assigned bank. Like Build, it also requires a Master file, but this file merely lists each .SAV to be included in the model, where each .SAV file name is given with the command iadd. Unlike Build, the Master file should not contain identities or other code. IdBuild is invoked from G7 by the command Model | IdBuild. This command also proceeds the compilation and linking of the model. (From DOS, IdBuild can be invoked by “idbuild master”.) The output files made by IdBuild are:

HIST.IND:A standard G bank that contains the series used or created by IdBuild. As in Build, variables on the right of fex commands are not included.
HEART.CPP:A compilable C++ program. Each of the iadd files becomes a separate, callable function with a name derived from the name of the iadd file from which it was created. It also contains a function (tserin) to read in all the time series from bws and make them accessible everywhere in the forecasting program. Both past and preliminary future values are available at all times.
HEART.H:Prototypes for the functions in the HEART.CPP file. This file makes it possible to call these functions from anywhere in the forecasting program.
TSERIES.INC:A file to be included in the main module of the forecasting program to declare the names of the variables in HIST.BNK as variables that can be used in the program.
CALLALL.CPP:A program to call all of the functions in HEART.CPP. It is used at the end of each year’s calculations to set or update rho adjustment factors.

The Master File for IdBuild

We will illustrate the forming of the master file with the Mudan model of China. The file for this model is:

bank cmdm
iadd invest.sav
iadd income.sav
iadd finance.sav
iadd pseudo.sav

Except for PSEUDO.SAV, the various .SAV files contain estimated equations. For example, INVEST.SAV begins

title Other State owned units investment
f sinvest = sibac$ + sirep$
r invn$35 = 40.346479*intercept + 0.110044*sinvest

The program recognizes PSEUDO.SAV as the name of a file with a special purpose. Namely, it puts into the model’s G bank those time series that are not needed in any of the code-image equations but are required elsewhere, perhaps in identities or detached coefficient equations. For Mudan, the beginning and end of the PSEUDO.SAV file are:

f trsa = trsa
f rpop = rpop
f upop = upop
f rscale = 1.0
f uscale = 1.0

These commands put the variables named trsa, rpop (rural population), and upop (urban population) into the model’s G bank. At the end, it initializes the rscale and uscale variables to 1.0 in all years. Most right-hand side expressions that are legal in G7 also would be legal here.

In the master file, the final q signals the end of input to IdBuild. IdBuild will produce from this command the following heart.cpp file:

#include <stdio.h>
#include "constant.h"
#include "matrix.h"
#include "dyme.h"
#include "groups.h"
#include "databank.h"
#include "vamfile.h"
#include "fixbank.h"
#include "dyme.ext"
#include "heart.h"
#include "tseries.ext"
FILE *fmatrix;
int i,j,k,err;
extern int t;
float depend;
/* end of standard prolog */

void investf(){
   /*  Other State owned units investment */
   sinvest[t]= sibac_[t]+ sirep_[t];
   /* invn$35 */ depend = 40.346479+0.110044* sinvest[t];
void incomef(){
void financef(){
void tserin(){"sibac$");"sirep$");"sinvest");

Note that all of the iadd files have been turned into functions whose names are formed by adding an ‘f’ to the end of the iadd file name. (Any $’s in the file name have been turned to _’s; there are none in the example.) What were f commands have become C statements with the exception that the ‘$’ character in variable names has been changed to an ‘_’. Examples of this change are seen in both the investf() function and the tserin() function. Any fex commands (none in the example) have disappeared, but the variable on the left of the fex has been created and entered into the data bank. Regression equations appear with the regression coefficients in the code. They calculate a variable, “depend”, which is passed to the routine modify(), along with the identification number of the dependent variable. The modify() routine then looks to see if there are any macro variable fixes – add, multiply, index, growth, skip, or rho adjustment – on that variable, and then stores the variable with the appropriate modification, if any. At the end of the HEART.CPP file is the tserin() function. It is called at the beginning of a run of the model to read into memory all the time-series variables. Finally, note that it contains all the variables which are in the PSEUDO.SAV file, even though no function was created by this file. This is the way to put into the data bank those variables, such as labfor, that appear in no code-image equation. The name “pseudo” is a keyword for the program; files by any other name create functions.

The HEART.H file created by IdBuild for this example is:

void investf();
void incomef();
void financef();
void exdgf();

It simply provides the prototypes required by C++ in any program that uses the functions in the HEART.CPP file.

The TSERIES.INC file is:

Tseries sibac_, sirep_, sinvest, invn_35, invn_36, d88, d90, invn_37,
rni, rpindex, rpop, rincome_, trsa, trsa_, invn_38, ulfi,

These files are “#included” in the forecasting program to declare that sibac_, sirep_, etc. are objects of the type “Tseries”. A “Tseries” is an object defined in the forecasting program that is designed to hold a time series. One of the things that a Tseries object “knows” how to do is to load itself from the assigned G data bank. Thus, in the tserin() function in the HEART.CPP program above, the command"ngdpc");

tells the ngdpc object to read in its data contents from the series called “ngdpc” in the data bank.

The final product of IdBuild is the CALLALL.CPP file, which for our example simply is:

#include "heart.h"
void callall(){

As its name suggests, callall() simply is a program to call all of the functions in the HEART.CPP file. Its function is in connection with rho adjustments, as will be seen in the forecasting program.

Combining Vector and Tseries Variables in a Function With IdBuild

When building interindustry macro models one usually needs to integrate the macro and the industry computations. For example, it often is necessary to form a macrovariable as a sum of components of a vector. Conversely, it may be that some sectoral variable is required on the right hand side of a macro equation. When this is the case, the IdBuild command isvector particularly is useful. This command indicates to IdBuild that a variable in one of the following save files is to be treated as an element of a vector, and not as a macrovariable or Tseries variable which is the default. For example, in the LIFT model of the U.S., the equation for railroad construction uses output of sector 59 (Railroads). Other equations use aggregates of output of many sectors. The included sections of files below show how this is handled. Here is part of the regression file for G7, CONSTR.REG:

save constr.sav
f outman = @csum(out,9 58)/1000.
f outbus = @csum(out,64,65,72,73,77 80)/1000.
f outtrade = @csum(out,69 71)/1000.
f outmin = @csum(out,2 6)/1000
ti 15. Railroad Construction
r cst15$ = cstoth, rpoil[2], rcbr[1], out59, doutrail, doutrail[1]
gr *

Here is a small MASTER file that will generate the code for this function only.

ba constr
isvector out,emp,pdm,cstk
iadd constr.sav

Here is the code for the constrf() function. Note that since the isvector command was in effect, IdBuild knows that “out” is a vector. Therefore, it passes “out” as one of the vectors in the argument list to constrf(), it writes out the csum function correctly as a method of type Vector, and it writes “out[59]” on the right hand side of the regression equation instead of “out59[t]”.

void constrf(Vector& out,Vector& emp,Vector& pdm,Vector& cstk){
   outman[t]=out.csum("9 58")/1000.;
   outbus[t]=out.csum("64,65,72,73,77 80")/1000.;
   outtrade[t]=out.csum("69 71")/1000.;
   outmin[t]=out.csum("2 6")/1000;
   /*  15. Railroad Construction */
   /* cst15_ */ depend =3699.702916+ 0.405094* cstoth[t]+5.812924* rpoil[t 2]+
   17.332162* rcbr[t 1]+0.020963* out[59]+0.040612* doutrail[t]+
   0.117446* doutrail[t 1];

At the present time, IdBuild doesn’t know how to handle lagged values of vector variables. In this case, you can make the vector variable a macrovariable and include code in your model to copy the macrovariable values to the vector and back again. For example, another regression in the construction equations mentioned above uses construction capital stock of category 16 lagged once (“cstk16[1]”). The way to handle this is as follows. Before opening the save file in G7, first do:

f cstk16$ = cstk16

Then, include cstk16$[1] on the right-hand side of the equation. In the simulation model, remember to fill the macrovariable with the value of construction capital stock of category 16 before calling the construction function:

cstk16_[t] = cstk[16];

Note that in addition to the iadd command, IdBuild has one more command not found in Build. That is the break command. Its format is:

break <filename>
After the break command, subsequent C++ code will go to the named file, with the extension .CPP appended, rather than to HEART.CPP. With a large model, HEART.CPP can become too large and may fail to compile. Also, somtimes it is more convenient to group the functions written by IdBuild into smaller logically organized files. This command may help in these cases.

Macrovariables also can be used on the right-hand side of equations for vector variables, called “detached-coefficient equations”. These are described in the InterDyme manual.

Please visit the Software pages of the Inforum web site for more details and to download the Fixer software:

Previous topic

Fixer: Fixing General Vectors and Matrices in InterDyme Models

This Page