#!/usr/bin/python3

import lxml
import lxml.etree
import ox
import os
import json
import sys

source = sys.argv[1]
target = '%s.srt' % os.path.splitext(source)[0]
fps = 25
data = []
tree = lxml.etree.parse(source)
'''
for g in tree.xpath('//clipitem//effect'):
    if g.findall('effectid')[0].text == "GraphicAndType":
        print (g.findall('name')[0].text)
'''

for g in tree.xpath('//clipitem'):
    if g.findall('name')[0].text != 'Graphic':
        continue
    #start/end = relative position of the clip in the parent sequence.
    #in/out indicate the portion of the source media file to reference.
    _in = int(g.findall('in')[0].text)
    _out = int(g.findall('out')[0].text)
    _start = int(g.findall('start')[0].text)
    _end = int(g.findall('end')[0].text)

    for efilter in g.findall('filter'):
        value = None
        for effect in efilter.findall('effect'):
            if effect.findall('effectid')[0].text == 'GraphicAndType':
                value = effect.findall('name')[0].text
                if not value:
                    print(lxml.etree.tostring(effect).decode())
                print(value)
        if _start == -1 and _end == -1:
            _start = _in
            _end = _out
        if _start == -1:
            _start = 0
        if value:
            print('i', _in, 'o', _out, '-> s', _start, 'e', _end, 'v', value)
            value = '\n'.join([v.strip() for v in value.strip().split('\r')])
            data.append({
                'in': _start/fps,
                'out': (_end-1)/fps,
                'value': value
            })
with open(target, 'wb') as f:
    f.write(ox.srt.encode(data))