add basic synthetic calibration test

test/test_synthetic_calibration.py

@@ -0,0 +1,48 @@
+import numpy as np
+import pytest
+
+from rascal.synthetic import SyntheticSpectrum
+from rascal.calibrator import Calibrator
+
+def test_default():
+
+    # Create a test spectrum with a simple linear relationship
+    # between pixels/wavelengths. The intercept is set to
+    # 100 nm and the dispersion is set to 2.
+    intercept = 100
+    dispersion = 2.0
+    s = SyntheticSpectrum(coefficients = [dispersion, intercept])
+
+    # We add a bunch of wavelegnths between 200-1200 nm
+    waves = np.linspace(200,1200, num=20)
+    peaks = s.get_pixels(waves)
+    assert len(peaks) > 0
+
+    # Set up the calibrator with the pixel values of our
+    # wavelengths
+    c = Calibrator(peaks=peaks)
+
+    # Arbitrarily we'll set the number of pixels to 768 (i.e.
+    # a max range of around 1500 nm
+    c.set_calibrator_properties(num_pix=768)
+
+    # Setup the Hough transform parameters
+    c.set_hough_properties(range_tolerance=100.,
+                           min_wavelength=100.,
+                           max_wavelength=1500.)
+
+    # Add our fake lines as the atlas
+    c.add_user_atlas(elements=["Test"]*len(waves), wavelengths=waves)
+    assert len(c.atlas) > 0
+
+    # And let's try and fit...
+    best_p, rms, residual, peak_utilisation = c.fit(max_tries=200)
+
+    best_p, x_fit, y_fit, residual, peak_utilisation = c.match_peaks(
+        best_p, refine=False, robust_refit=True)
+
+    fit_diff = c.polyval(x_fit, best_p) - y_fit
+    rms = np.sqrt(np.sum(fit_diff**2 / len(x_fit)))
+
+    assert peak_utilisation > 0.9
+    assert rms < 1e-6