Add new formatters.py module

This commit is contained in:
Chris Howell 2020-09-01 15:21:47 -07:00
parent 7a47fc83ad
commit f3dc6f508f
1 changed files with 91 additions and 0 deletions

View File

@ -0,0 +1,91 @@
import json
class Formatter(object):
"""Formatter should be used as an abstract base class.
Formatter classes should inherit from this class and implement
their own .format() method which should return a string. A
transcript is represented by a List of Dictionary items.
:param transcript: list representing 1 or more transcripts
:type transcript: list
"""
def __init__(self, transcript):
if not isinstance(transcript, list):
raise TypeError("'transcript' must be of type: List")
self._transcript = transcript
def format(self, **kwargs):
raise NotImplementedError('A subclass of Formatter must implement ' \
'their own .format() method.')
class JSONFormatter(Formatter):
def format(self, **kwargs):
"""Converts a transcript into a JSON string.
:return: A JSON string representation of the transcript.'
:rtype str
"""
return json.dumps(self._transcript, **kwargs)
class TextFormatter(Formatter):
def format(self, **kwargs):
"""Converts a transcript into plain text with no timestamps.
:return: all transcript text lines separated by newline breaks.'
:rtype str
"""
return "\n".join(line['text'] for line in self._transcript)
class WebVTTFormatter(Formatter):
def _seconds_to_timestamp(self, time):
"""Helper that converts `time` into a transcript cue timestamp.
:reference: https://www.w3.org/TR/webvtt1/#webvtt-timestamp
:param time: a float representing time in seconds.
:type time: float
:return: a string formatted as a cue timestamp, 'HH:MM:SS.MS'
:rtype str
:example:
>>> self._seconds_to_timestamp(6.93)
'00:00:06.930'
"""
time = float(time)
hours, mins, secs = (
int(time) // 3600,
int(time) // 60,
int(time) % 60,
)
ms = int(round((time - int(time))*1000, 2))
return "{:02d}:{:02d}:{:02d}.{:03d}".format(hours, mins, secs, ms)
def format(self, **kwargs):
"""A basic implementation of WEBVTT formatting.
:reference: https://www.w3.org/TR/webvtt1/#introduction-caption
"""
lines = []
for i, line in enumerate(self._transcript):
if i < len(self._transcript)-1:
# Looks ahead, use next start time since duration value
# would create an overlap between start times.
time_text = "{} --> {}".format(
self._seconds_to_timestamp(line['start']),
self._seconds_to_timestamp(self._transcript[i+1]['start'])
)
else:
# Reached the end, cannot look ahead, use duration now.
duration = line['start'] + line['duration']
time_text = "{} --> {}".format(
self._seconds_to_timestamp(line['start']),
self._seconds_to_timestamp(duration)
)
lines.append("{}\n{}".format(time_text, line['text']))
return "WEBVTT\n\n" + "\n\n".join(lines) + "\n"