def _check_empty_line(fh):
"""Check that the next line in `fh` is empty or whitespace-only."""
line = next(fh, None)
if line is None:
raise OrdinationFormatError(
"Reached end of file while looking for blank line separating "
"sections.")
if line.strip():
raise OrdinationFormatError("Expected an empty line.")
def _parse_header(fh, header_id, num_dimensions):
line = next(fh, None)
if line is None:
raise OrdinationFormatError(
"Reached end of file while looking for %s header." % header_id)
header = line.strip().split('\t')
# +1 for the header ID
if len(header) != num_dimensions + 1 or header[0] != header_id:
raise OrdinationFormatError("%s header not found." % header_id)
return header
def _parse_array_section(fh, header_id, has_ids=True):
"""Parse an array section of `fh` identified by `header_id`."""
# Parse the array header
header = _parse_header(fh, header_id, 2)
# Parse the dimensions of the array
rows = int(header[1])
cols = int(header[2])
ids = None
if rows == 0 and cols == 0:
# The ordination method didn't generate the array data for 'header', so
# set it to None
data = None
elif rows == 0 or cols == 0:
# Both dimensions should be 0 or none of them are zero
raise OrdinationFormatError("One dimension of %s is 0: %d x %d" %
(header_id, rows, cols))
else:
# Parse the data
data = np.empty((rows, cols), dtype=np.float64)
if has_ids:
ids = []
for i in range(rows):
# Parse the next row of data
line = next(fh, None)
if line is None:
raise OrdinationFormatError(
"Reached end of file while looking for row %d in %s "
"section." % (i + 1, header_id))
vals = line.strip().split('\t')
if has_ids:
ids.append(vals[0])
vals = vals[1:]
if len(vals) != cols:
raise OrdinationFormatError(
"Expected %d values, but found %d in row %d." %
(cols, len(vals), i + 1))
data[i, :] = np.asarray(vals, dtype=np.float64)
data = pd.DataFrame(data, index=ids)
return data
def _check_length_against_eigvals(data, eigvals, label):
if data is not None:
num_vals = data.shape[-1]
num_eigvals = eigvals.shape[-1]
if num_vals != num_eigvals:
raise OrdinationFormatError(
"There should be as many %s as eigvals: %d != %d" %
(label, num_vals, num_eigvals))
def _parse_vector_section(fh, header_id):
header = _parse_header(fh, header_id, 1)
# Parse how many values we are waiting for
num_vals = int(header[1])
if num_vals == 0:
# The ordination method didn't generate the vector, so set it to None
vals = None
else:
# Parse the line with the vector values
line = next(fh, None)
if line is None:
raise OrdinationFormatError(
"Reached end of file while looking for line containing values "
"for %s section." % header_id)
vals = np.asarray(line.strip().split('\t'), dtype=np.float64)
if len(vals) != num_vals:
raise OrdinationFormatError(
"Expected %d values in %s section, but found %d." %
(num_vals, header_id, len(vals)))
return vals
def _ordination_to_ordination_results(fh):
eigvals = _parse_vector_section(fh, 'Eigvals')
if eigvals is None:
raise OrdinationFormatError("At least one eigval must be present.")
_check_empty_line(fh)
prop_expl = _parse_vector_section(fh, 'Proportion explained')
_check_length_against_eigvals(prop_expl, eigvals,
'proportion explained values')
_check_empty_line(fh)
species, species_ids = _parse_array_section(fh, 'Species')
_check_length_against_eigvals(species, eigvals, 'coordinates per species')
_check_empty_line(fh)
site, site_ids = _parse_array_section(fh, 'Site')
_check_length_against_eigvals(site, eigvals, 'coordinates per site')
_check_empty_line(fh)
# biplot does not have ids to parse (the other arrays do)
biplot, _ = _parse_array_section(fh, 'Biplot', has_ids=False)
_check_empty_line(fh)
cons, cons_ids = _parse_array_section(fh, 'Site constraints')
if cons_ids is not None and site_ids is not None:
if cons_ids != site_ids:
raise OrdinationFormatError(
"Site constraints ids and site ids must be equal: %s != %s" %
(cons_ids, site_ids))
return OrdinationResults(eigvals=eigvals,
species=species,
site=site,
biplot=biplot,
site_constraints=cons,
proportion_explained=prop_expl,
species_ids=species_ids,
site_ids=site_ids)
def _ordination_to_ordination_results(fh):
eigvals = _parse_vector_section(fh, 'Eigvals')
if eigvals is None:
raise OrdinationFormatError("At least one eigval must be present.")
_check_empty_line(fh)
prop_expl = _parse_vector_section(fh, 'Proportion explained')
_check_length_against_eigvals(prop_expl, eigvals,
'proportion explained values')
_check_empty_line(fh)
species = _parse_array_section(fh, 'Species')
_check_length_against_eigvals(species, eigvals,
'coordinates per species')
_check_empty_line(fh)
site = _parse_array_section(fh, 'Site')
_check_length_against_eigvals(site, eigvals,
'coordinates per site')
_check_empty_line(fh)
# biplot does not have ids to parse (the other arrays do)
biplot = _parse_array_section(fh, 'Biplot', has_ids=False)
_check_empty_line(fh)
cons = _parse_array_section(fh, 'Site constraints')
if cons is not None and site is not None:
if not np.array_equal(cons.index, site.index):
raise OrdinationFormatError(
"Site constraints ids and site ids must be equal: %s != %s" %
(cons.index, site.index))
return OrdinationResults(
short_method_name='', long_method_name='', eigvals=eigvals,
features=species, samples=site, biplot_scores=biplot,
sample_constraints=cons, proportion_explained=prop_expl)