Predict requires DataFrame that contains the predicted variable in 3.1.0

[smpawlowski]

Describe the bug LinearModel.predict requires DataFrame that contains the predicted variable. This likely affects other models as well.

Expected behavior Predict should work with explanatory variables only.

Actual behavior Predict throws when predicted variable is not included in the DataFrame.

Code snippet

@Test
public void test_formula_OLS() {
    double[] x = {1, 2, 3};
    double[] y = {1, 2, 3};
    DataFrame df = DataFrame.of(DoubleVector.of("x", x),
                                DoubleVector.of("y", y));
    LinearModel regr = OLS.fit(Formula.lhs("y"), df);

    double[] x_pred = {4, 5, 6};
    double[] y_pred = regr.predict(DataFrame.of( DoubleVector.of("x", x_pred)));
    for(int i=0; i<x_pred.length; i++) {
        Assert.assertEquals(x_pred[i], y_pred[i], 1e-9);
    }
}

Exception

java.lang.IllegalArgumentException: Field y doesn't exist

at smile.data.type.StructType.indexOf(StructType.java:103)
at smile.data.formula.Variable$1.<init>(Variable.java:80)
at smile.data.formula.Variable.bind(Variable.java:78)
at smile.data.formula.Formula.bind(Formula.java:360)
at smile.data.formula.Formula.x(Formula.java:497)
at smile.data.formula.Formula.matrix(Formula.java:546)
at smile.regression.LinearModel.predict(LinearModel.java:358)
at models.TestSmileRegression.test_formula_OLS(TestSmileRegression.java:22)

Input data In the snippet ^

Additional context

• Sample code throws in ver 3.1.0, but works fine in 2.6.0

[haifengl]

Thanks for reporting. The fix is in the master branch now.

[haifengl]

Please let's know if it address your issues. If so, we will make a new release. Thanks!

[smpawlowski]

Hi! The code and test update look good. Unfortunately I can't fully test at the moment because of firewall/source code restrictions at my company. I can easily test if you make a release. Thanks for looking into it so quickly!

[haifengl]

v3.1.1 is released.

[smpawlowski]

3.1.1 fixes the issue! Thanks again!