standardize test_data before predict/decision_function in SVM

As far as I know, when we standardize/normalize training_data, we should do the test_data as well as. I am looking at some codes and became confused when it has been doing here:

scaler = preprocessing.StandardScaler()
clf = svm.LinearSVC(class_weight='balanced',C=self.C, max_iter=max_iter)
pipeline = Pipeline([('scaler',scaler),('svm',clf)])
fit = pipeline.fit(design_matrix, classes)
self.fit = fit
...
# no additional prepossess on feature_array 
predicted_classes = self.fit.decision_function(feature_array)

could u please clarify it for me?