def _nearest(scdf, ocdf, suffix="_b", how=None, overlap=True, **kwargs):
if overlap:
nearest_df, df_to_find_nearest_in = _overlapping_for_nearest(
scdf, ocdf, suffix, **kwargs)
else:
df_to_find_nearest_in = scdf
df_to_find_nearest_in = sort_one_by_one(df_to_find_nearest_in, "Start",
"End")
ocdf = sort_one_by_one(ocdf, "Start", "End")
df_to_find_nearest_in.index = pd.Index(range(len(df_to_find_nearest_in)))
if how == "next":
r_idx, dist = _next_nonoverlapping(df_to_find_nearest_in.End,
ocdf.Start, ocdf.index.values)
elif how == "previous":
r_idx, dist = _previous_nonoverlapping(df_to_find_nearest_in.Start,
ocdf.End)
else:
previous_r_idx, previous_dist = _previous_nonoverlapping(
df_to_find_nearest_in.Start, ocdf.End)
next_r_idx, next_dist = _next_nonoverlapping(df_to_find_nearest_in.End,
ocdf.Start,
ocdf.index.values)
r_idx, dist = nearest_nonoverlapping(previous_r_idx, previous_dist,
next_r_idx, next_dist)
ocdf = ocdf.reindex(
r_idx,
fill_value=-1) # instead of np.nan, so ints are not promoted to float
ocdf.index = df_to_find_nearest_in.index
ocdf.insert(ocdf.shape[1], "Distance",
pd.Series(dist, index=ocdf.index).fillna(-1).astype(int))
r_idx = pd.Series(r_idx, index=ocdf.index)
df_to_find_nearest_in = df_to_find_nearest_in.drop(
r_idx.loc[r_idx == -1].index)
df = df_to_find_nearest_in.join(ocdf, rsuffix=suffix)
if overlap and not df.empty and not nearest_df.empty:
df = pd.concat([nearest_df, df])
elif overlap and not nearest_df.empty:
df = nearest_df
df = df.drop("Chromosome" + suffix, axis=1)
return df
def _nearest(self, other, strandedness, suffix="_b", how=None, overlap=True):
if overlap:
nearest_df, df_to_find_nearest_in = _overlapping_for_nearest(self, other, strandedness, suffix)
else:
df_to_find_nearest_in = self.df
other_strand = {"+": "-", "-": "+"}
if self.stranded and strandedness: # chromosome and strand
grpby_key = "Chromosome Strand".split()
else:
grpby_key = "Chromosome"
other_dfs = {k: d for k, d in other.df.groupby(grpby_key)}
dfs = []
for key, scdf in df_to_find_nearest_in.groupby(grpby_key):
if len(key) == 2 and strandedness == "opposite":
other_key = key[0], other_strand[key[1]]
else:
other_key = key
if not other_key in other_dfs:
continue
ocdf = other_dfs[other_key]
scdf.index = pd.Index(range(len(scdf)))
if how == "next":
r_idx, dist = _next_nonoverlapping(scdf.End, ocdf.Start, ocdf.index.values)
elif how == "previous":
r_idx, dist = _previous_nonoverlapping(scdf.Start, ocdf.End, ocdf.index.values)
else:
previous_r_idx, previous_dist = _previous_nonoverlapping(scdf.Start, ocdf.End, ocdf.index.values)
next_r_idx, next_dist = _next_nonoverlapping(scdf.End, ocdf.Start, ocdf.index.values)
r_idx, dist = nearest_nonoverlapping(previous_r_idx,
previous_dist,
next_r_idx, next_dist)
ocdf = ocdf.reindex(r_idx, fill_value=-1) # instead of np.nan, so ints are not promoted to float
ocdf.index = scdf.index
ocdf.insert(ocdf.shape[1], "Distance", pd.Series(dist, index=ocdf.index).fillna(-1).astype(int))
ocdf.drop("Chromosome", axis=1, inplace=True)
r_idx = pd.Series(r_idx, index=ocdf.index)
scdf = scdf.drop(r_idx.loc[r_idx == -1].index)
result = scdf.join(ocdf, rsuffix=suffix)
dfs.append(result)
if dfs:
df = pd.concat(dfs)
else:
df = pd.DataFrame(columns="Chromosome Start End Strand".split())
if overlap and not df.empty and not nearest_df.empty:
df = pd.concat([nearest_df, df])
elif overlap and not nearest_df.empty:
df = nearest_df
return df
def _nearest(scdf, ocdf, kwargs):
if scdf.empty or ocdf.empty:
return None
overlap = kwargs["overlap"]
how = kwargs["how"]
suffix = kwargs["suffix"]
if how == "upstream":
strand = scdf.Strand.iloc[0]
how = {"+": "previous", "-": "next"}[strand]
elif how == "downstream":
strand = scdf.Strand.iloc[0]
how = {"+": "next", "-": "previous"}[strand]
ocdf = ocdf.reset_index(drop=True)
if overlap:
nearest_df, df_to_find_nearest_in = _overlapping_for_nearest(
scdf, ocdf, suffix)
else:
df_to_find_nearest_in = scdf
if not df_to_find_nearest_in.empty:
df_to_find_nearest_in = sort_one_by_one(df_to_find_nearest_in, "Start",
"End")
ocdf = sort_one_by_one(ocdf, "Start", "End")
df_to_find_nearest_in.index = pd.Index(
range(len(df_to_find_nearest_in)))
if how == "next":
r_idx, dist = _next_nonoverlapping(df_to_find_nearest_in.End,
ocdf.Start, ocdf.index.values)
elif how == "previous":
r_idx, dist = _previous_nonoverlapping(df_to_find_nearest_in.Start,
ocdf.End)
else:
previous_r_idx, previous_dist = _previous_nonoverlapping(
df_to_find_nearest_in.Start, ocdf.End)
next_r_idx, next_dist = _next_nonoverlapping(
df_to_find_nearest_in.End, ocdf.Start, ocdf.index.values)
r_idx, dist = nearest_nonoverlapping(previous_r_idx, previous_dist,
next_r_idx, next_dist)
ocdf = ocdf.reindex(r_idx)
ocdf.index = df_to_find_nearest_in.index
ocdf = _insert_distance(ocdf, dist, suffix)
r_idx = pd.Series(r_idx, index=ocdf.index)
df_to_find_nearest_in = df_to_find_nearest_in.drop(
r_idx.loc[r_idx == -1].index)
df = df_to_find_nearest_in.join(ocdf, rsuffix=suffix)
if overlap and "df" in locals() and not df.empty and not nearest_df.empty:
df = pd.concat([nearest_df, df], sort=False)
elif overlap and not nearest_df.empty:
df = nearest_df
df = df.drop("Chromosome" + suffix, axis=1)
return df