Merge pull request #39 from computational-metabolomics/dev

Add a number of bug fixes and improvements

.travis.yml

@@ -1,6 +1,7 @@
 language: python
 python:
   - "2.7"
+
 install:
   - sudo apt-get update
   - if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then
@@ -15,9 +16,13 @@ install:
   - conda update -q conda
   - conda info -a
 
-  - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION numpy=1.13.0 scipy=0.19.1 pymzml=0.7.8 pythonnet=2.3.0 h5py=2.7.0 fastcluster=1.1.23 -c conda-forge -c bioconda
+  - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION numpy=1.13.0 scipy=1.0 pymzml=0.7.10 pythonnet=2.3.0 h5py=2.7.0 fastcluster=1.1.23 coverage green codecov -c conda-forge -c bioconda
   - source activate test-environment
   - python setup.py install
 
 script:
+  - dimspy --help
   - python setup.py test
+
+after_script:
+  - python -m codecov

README.rst

@@ -1,6 +1,6 @@
 DIMSpy
 ======
-|Version| |Py versions| |Git| |Bioconda| |Build Status (Travis)| |Build Status (AppVeyor)| |License| |RTD doc|
+|Version| |Py versions| |Git| |Bioconda| |Build Status (Travis)| |Build Status (AppVeyor)| |License| |RTD doc| |codecov|
 
 Python package to process direct-infusion mass spectrometry-based metabolomics and lipidomics data
 
@@ -64,3 +64,6 @@ Released under the GNU General Public License v3.0 (see `LICENSE file <https://g
 
 .. |RTD doc| image:: https://img.shields.io/badge/documentation-RTD-71B360.svg?style=flat&maxAge=3600
    :target: https://computational-metabolomics.github.io/dimspy/
+
+.. |codecov| image:: https://codecov.io/gh/computational-metabolomics/dimspy/branch/master/graph/badge.svg
+   :target: https://codecov.io/gh/computational-metabolomics/dimspy

dimspy/__main__.py

@@ -6,7 +6,7 @@
 import h5py
 import tools
 from portals import hdf5_portal
-from . import __version__
+from dimspy import __version__
 
 
 def map_delimiter(delimiter):

dimspy/models/peaklist.py

@@ -99,6 +99,9 @@ def __len__(self):
     def __str__(self):
         return self.to_str(', ')
 
+    def __repr__(self):
+        return self.to_str('\t')
+
     def __getitem__(self, item):
         if type(item) in (int, slice, list, np.ndarray):
             return self.get_peak(item)
@@ -348,7 +351,7 @@ def add_attribute(self, attr_name, attr_value, attr_dtype=None, is_flag=False,
         attr_name = str(attr_name) # rfn.append_fields doesn't recognise unicode
 
         adt = bool if is_flag else \
-              attr_dtype if attr_dtype is not None else \
+              attr_dtype if attr_dtype not in (None, str, unicode) else \
               attr_value.dtype.str if hasattr(attr_value, 'dtype') else \
               ('S%d' % max(map(len, attr_value))) if type(attr_value[0]) in (unicode, str) else \
               type(attr_value[0])

dimspy/portals/thermo_raw_portal.py

@@ -61,7 +61,7 @@ def peaklist(self, scan_id, function_noise="noise_packets"):
         else:
             mzs, ints, baseline, noise = [], [], [], []
 
-        if function_noise == "noise_packets":
+        if function_noise == "noise_packets" and len(ints) > 0:
             snr = [p.SignalToNoise for p in scan.GetCentroids()]
         elif function_noise == "median" and len(ints) > 0:
             snr = ints / np.median(ints)

dimspy/process/peak_alignment.py

@@ -31,21 +31,34 @@ def _cluster_peaks(mzs, ppm, distype='euclidean', linkmode='centroid'):
     if len(mzs) == 1:
         return np.zeros_like(mzs, dtype=int).reshape((-1, 1))
 
+    outer_mzs = np.add.outer(mzs, mzs)
+    np.fill_diagonal(outer_mzs, 0)
+
+    # avg_mz_pair = np.divide(outer_mzs, 2)
+    outer_mzs /= 2  # inplace operation to reduce memory usage
+
+    # mdist_mz_pair = squareform(avg_mz_pair)
+    mdist_mz_pair = squareform(outer_mzs)
+    del outer_mzs  # reduce memory use
+
     m = np.column_stack([mzs])
     mdist = fc.pdist(m, metric=distype)
+    del m
 
-    outer_mzs = np.add.outer(mzs, mzs)
-    np.fill_diagonal(outer_mzs, 0)
-    avg_mz_pair = np.divide(outer_mzs, 2)
-    mdist_mz_pair = squareform(avg_mz_pair)
-    relative_errors = np.multiply(mdist_mz_pair, 1e-6)
+    # relative_errors = np.multiply(mdist_mz_pair, 1e-6)
+    mdist_mz_pair *= 1e-6  # inplace operation to reduce memory usage
 
     with np.errstate(divide='ignore', invalid='ignore'):  # using errstate context to avoid seterr side effects
-        m_mass_tol = np.divide(mdist, relative_errors)
-        m_mass_tol[np.isnan(m_mass_tol)] = 0.0
-    z = fc.linkage(m_mass_tol, method=linkmode)
+        # m_mass_tol = np.divide(mdist, relative_errors)
+        mdist /= mdist_mz_pair  # inplace operation to reduce memory usage
+        # m_mass_tol[np.isnan(m_mass_tol)] = 0.0
+        mdist[np.isnan(mdist)] = 0.0
+
+    # z = fc.linkage(m_mass_tol, method=linkmode)
+    z = fc.linkage(mdist, method=linkmode)
+    del mdist, mdist_mz_pair
 
-    # cut tree at ppm threshold & order matches the order of mzs
+    # cut tree at ppm threshold
     return cluster.hierarchy.cut_tree(z, height=ppm)
 
 

requirements.txt

@@ -2,6 +2,6 @@ numpy==1.13.3
 scipy==1.0.0
 fastcluster==1.1.23
 plotly==2.5.1
-pymzml==0.7.8
+pymzml==0.7.10
 h5py==2.7.0
-pythonnet==2.3.0
+pythonnet>=2.3.0