Link directly to PDF files; remove symlinks.

doc/study.rst

@@ -121,7 +121,7 @@ where the :meth:`Table.write_asgl`
 method writes out an ASGL script and the MDT data in a form suitable for
 plotting (which we then execute with ASGL using Python's
 :meth:`os.system` method). This gives an
-:download:`impact of resolution plot <html/pdf/impact-of-resolution.pdf>`.
+:download:`impact of resolution plot <../constr2005/impact-of-resolution/asgl2-a.pdf>`.
 
 
 .. _chembonds:
@@ -142,8 +142,8 @@ length. The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/bonds/asgl.py
    :language: python
 
-giving :download:`a set of bond plots <html/pdf/bonds.pdf>`. Notice that
-here we use the :meth:`Table.reshape` method,
+giving :download:`a set of bond plots <../constr2005/bonds/asgl1-a.pdf>`.
+Notice that here we use the :meth:`Table.reshape` method,
 which can reshape a table by reordering the features, and/or reducing the bin
 ranges (offset or shape) of these features. In this case we don't change the
 feature order, or the offset (leaving it at the default 0,0,0) but we do
@@ -197,7 +197,7 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/angles/asgl.py
    :language: python
 
-giving :download:`a set of angle plots <html/pdf/angles.pdf>`.
+giving :download:`a set of angle plots <../constr2005/angles/asgl1-a.pdf>`.
 
 The situation is similar to that for the chemical bonds, except that there are
 also four cases of bi-modal (as opposed to mono-modal) distributions:
@@ -229,7 +229,8 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/impropers/asgl.py
    :language: python
 
-giving :download:`a set of improper plots <html/pdf/impropers.pdf>`.
+giving
+:download:`a set of improper plots <../constr2005/impropers/asgl1-a.pdf>`.
 
 The final MODELLER MDT library is produced with:
 
@@ -254,7 +255,8 @@ and
 .. literalinclude:: ../constr2005/chi1/impact-of-resolution/asgl.py
    :language: python
 
-giving :download:`this output <html/pdf/chi1-impact.pdf>`
+giving
+:download:`this output <../constr2005/chi1/impact-of-resolution/asgl2-a.pdf>`
 which clearly shows that X-ray structures at resolution of at least 2.0 Å
 are just fine. X-ray structures above that resolution and NMR structures
 (whose resolution is set artificially to 0.45 Å for the purposes of MDT
@@ -273,7 +275,7 @@ and the contents then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/chi1/asgl.py
    :language: python
 
-giving :download:`a set of χ1 plots <html/pdf/chi1.pdf>`.
+giving :download:`a set of χ1 plots <../constr2005/chi1/asgl1-a.pdf>`.
 
 The weights, means, and standard deviations of the Gaussians were obtained
 by least-squares fitting with ASGL (with the script below) and are manually
@@ -300,7 +302,7 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/chi2/asgl.py
    :language: python
 
-giving :download:`a set of χ2 plots <html/pdf/chi2.pdf>`.
+giving :download:`a set of χ2 plots <../constr2005/chi2/asgl1-a.pdf>`.
 
 The final MODELLER MDT library is produced with:
 
@@ -334,7 +336,7 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/chi3/asgl.py
    :language: python
 
-giving :download:`a set of χ3 plots <html/pdf/chi3.pdf>`.
+giving :download:`a set of χ3 plots <../constr2005/chi3/asgl1-a.pdf>`.
 The final MODELLER MDT library is produced with:
 
 .. literalinclude:: ../constr2005/chi3/modlib.py
@@ -359,7 +361,7 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/chi4/asgl.py
    :language: python
 
-giving :download:`a set of χ4 plots <html/pdf/chi4.pdf>`.
+giving :download:`a set of χ4 plots <../constr2005/chi4/asgl1-a.pdf>`.
 The final MODELLER MDT library is produced with:
 
 .. literalinclude:: ../constr2005/chi4/modlib.py
@@ -384,7 +386,7 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/phi/asgl.py
    :language: python
 
-giving :download:`a set of Φ plots <html/pdf/phi.pdf>`.
+giving :download:`a set of Φ plots <../constr2005/phi/asgl1-a.pdf>`.
 The final MODELLER MDT library is produced with:
 
 .. literalinclude:: ../constr2005/phi/modlib.py
@@ -409,7 +411,7 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/psi/asgl.py
    :language: python
 
-giving :download:`a set of Ψ plots <html/pdf/psi.pdf>`.
+giving :download:`a set of Ψ plots <../constr2005/psi/asgl1-a.pdf>`.
 The final MODELLER MDT library is produced with:
 
 .. literalinclude:: ../constr2005/psi/modlib.py
@@ -440,13 +442,14 @@ distributions with finer sampling of 0.5°:
 .. literalinclude:: ../constr2005/omega/make-mdt.py
    :language: python
 
-The :download:`distribution in raw form <html/pdf/omega.pdf>`
+The :download:`distribution in raw form <../constr2005/omega/asgl1-a.pdf>`
 is then plotted with:
 
 .. literalinclude:: ../constr2005/omega/asgl.py
    :language: python
 
-and :download:`in logarithmic form <html/pdf/omega-log.pdf>` with:
+and :download:`in logarithmic form <../constr2005/omega/asgl2-a.pdf>`
+with:
 
 .. literalinclude:: ../constr2005/omega/asgl-log.py
    :language: python
@@ -482,7 +485,8 @@ The contents of the MDT table are then plotted with ASGL as follows:
 .. literalinclude:: ../constr2005/omega/weights/asgl.py
    :language: python
 
-giving :download:`an omega weights plot <html/pdf/omega-weights.pdf>`.
+giving
+:download:`an omega weights plot <../constr2005/omega/weights/asgl1-a.pdf>`.
 
 The library :file:`omega.py` is edited manually to replace the
 means and standard deviations with `179.8  0.0   2.3  2.3`.
@@ -508,7 +512,7 @@ They are plotted with
 .. literalinclude:: ../constr2005/phipsi/asgl.py
    :language: python
 
-giving :download:`a set of Φ/Ψ plots <html/pdf/phipsi.pdf>`.
+giving :download:`a set of Φ/Ψ plots <../constr2005/phipsi/asgl1-a.pdf>`.
 
 The distributions are clearly not 2D Gaussian functions and need to be
 approximated by 2D cubic splines. Exploring and visualizing various smoothing