old thing from 2013

runs.py

import pandas as pd
import numpy as np

def _run_starts_ends(bools):
    """
    Returns a start and end bool array that marks the endpoints of 
    each contigious group of Trues

    Parameters
    ----------
    bools : ndarray
        Boolean array
    """
    diffs = bools.astype(int).diff()
    starts = diffs == 1
    ends = diffs == -1
    # diff won't catch inital True
    if bools.iloc[0] == True:
        starts.iloc[0] = True
    return starts, ends

def run_lengths(bools):
    """
    Returns a Series marking the length of each run at its first period. 

    Parameters
    ----------
    bools : ndarray
        Boolean array
    
    Returns
    -------
    runs : pd.Series
        Marks the beginning of each run with the length of run. Uses same index
        as bools
    """
    starts, ends = _run_starts_ends(bools)

    # get the pos-posex of true values
    start_pos = list(np.where(starts)[0])
    end_pos = list(np.where(ends)[0])

    if bools.iloc[-1] == True:
        end_pos.append(len(bools))

    start_pos = np.array(start_pos)
    end_pos = np.array(end_pos)

    run_lengths = end_pos - start_pos
    runs = pd.Series(np.nan, index=bools.index)
    runs.put(start_pos, run_lengths)
    return runs

def break_lengths(bools):
    """
    """
    breaks = run_lengths(~bools)
    return breaks

def grab_runs(bools):
    """
    """
    breaks = break_lengths(bools)
    runs = pd.DataFrame({'run_length': run_lengths(bools).dropna()})
    first = runs.index[0]
    last = runs.index[-1]
    breaks_before = pd.Series(np.nan, index=runs.index)
    breaks_after = pd.Series(np.nan, index=runs.index)
    bb = breaks.ix[:last].dropna() # breaks_before
    ba = breaks.ix[first:].dropna() # break_after
    breaks_before[len(runs)-len(bb):] = bb
    breaks_after[:len(ba)] = ba

    runs['breaks_after'] = breaks_after
    runs['breaks_before'] = breaks_before

    return runs