From a2d3a9d5cc1110174537202ea81f419a68130a9f Mon Sep 17 00:00:00 2001
From: tangstad <tangstad@pvv.ntnu.no>
Date: Fri, 28 Apr 2006 11:44:55 +0000
Subject: [PATCH] Made PCA more generic, even enable PCA of existing
 PCA-results as dimension names are modified to handle dimension name
 component in input.

---
 system/plots.py            |  6 ++++--
 workflows/affy_workflow.py | 25 ++++++++++++++++++-------
 2 files changed, 22 insertions(+), 9 deletions(-)

diff --git a/system/plots.py b/system/plots.py
index 079b346..31f42c2 100644
--- a/system/plots.py
+++ b/system/plots.py
@@ -392,9 +392,11 @@ class LinePlot(Plot):
         rows, cols = self._bg_matrix.shape
         self.ax.imshow(self._bg_matrix, cmap=cm.Greys, extent=(0.5, cols+0.5, self._ymin, self._ymax))
 
+        dim_2, dim_1 = self._dataset.get_dim_names()
+
         if selection:
-            ids = selection['ids'] # current identifiers
-            index = [ind for id,ind in self._dataset['ids'].items() if id in ids] #conversion to index
+            ids = selection[dim_2] # current identifiers
+            index = [ind for id,ind in self._dataset[dim_2].items() if id in ids] #conversion to index
             for i in index:
                 line = self._dataset.get_matrix()[i]
                 self.ax.plot(range(1, len(line)+1), line)
diff --git a/workflows/affy_workflow.py b/workflows/affy_workflow.py
index 445abee..89a4dc4 100644
--- a/workflows/affy_workflow.py
+++ b/workflows/affy_workflow.py
@@ -3,6 +3,7 @@ from system import dataset, logger, plots, workflow
 from scipy import randn
 import cPickle
 
+
 class AffyWorkflow (workflow.Workflow):
 
     name = 'Affy Workflow'
@@ -34,7 +35,7 @@ class TestDataFunction(workflow.Function):
         logger.log('notice', 'Injecting foo test data')
         x = randn(20,30)
         X = dataset.Dataset(x)
-        return [X, plots.SinePlot()]
+        return [X, plots.LinePlot(X)]
 
 
 class DatasetLoadFunction(workflow.Function):
@@ -153,22 +154,32 @@ class PCAFunction(workflow.Function):
         import rpy
         
         dim_2, dim_1 = data.get_dim_names()
-        
     
         silent_eval = rpy.with_mode(rpy.NO_CONVERSION, rpy.r)
         rpy.with_mode(rpy.NO_CONVERSION, rpy.r.assign)("m", data.get_matrix())
         silent_eval("t = prcomp(t(m))")
 
+        # we make a unique name for component dimension
+        c = 0
+        component_dim = prefix = "component"
+        while component_dim in data.get_all_dims():
+            component_dim = prefix + "_" + str(c)
+            c += 1
+
         T_ids = map(str, range(1, rpy.r("dim(t$x)")[1]+1))
         T = dataset.Dataset(rpy.r("t$x"), [(dim_1, data.get_identifiers(dim_1)),
-                                   ("component", T_ids)], name="T")
+                                           (component_dim, T_ids)],
+                            all_dims = data.get_all_dims(), name="T")
         P = dataset.Dataset(rpy.r("t$rotation"), [(dim_2, data.get_identifiers(dim_2)),
-                                          ("component", T_ids)], name="P")
-
+                                                  (component_dim, T_ids)],
+                            all_dims = data.get_all_dims(), name="P")
         # cleanup
         rpy.r.rm(["t", "m"])
 
-        loading_plot = plots.ScatterPlot(P,'ids','component','1','2', "Loadings")
-        score_plot = plots.ScatterPlot(T,'filename','component','1','2', "Scores")
+        loading_plot = plots.ScatterPlot(P, dim_2, component_dim, '1', '2',
+                                         "Loadings")
+        score_plot = plots.ScatterPlot(T, dim_1,component_dim, '1', '2',
+                                       "Scores")
         
         return [T, P, loading_plot, score_plot]
+