---
title: "Parsing subtitles in srt format"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Parsing subtitles in srt format}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = FALSE,
  comment = "#>"
)
```

```{r setup}
library(flexo)
```

## Parsing subtitles in srt format

The [srt](https://www.speechpad.com/captions/srt) subtitle format is a 
simple representation of subtitles for video which consists of timestamped
lines of text.


# SRT format example

The first 10 lines of dialogue from "It's a Wonderful Life" in `srt` format

```{r}
srt <- "
1
00:01:25,210 --> 00:01:28,004
I owe everything to George Bailey.

2
00:01:28,422 --> 00:01:30,298
Help him, dear Father.

3
00:01:30,674 --> 00:01:33,718
Joseph, Jesus and Mary,

4
00:01:33,802 --> 00:01:36,429
help my friend Mr. Bailey.

5
00:01:36,889 --> 00:01:39,515
Help my son George tonight.

6
00:01:40,350 --> 00:01:42,226
He never thinks about himself, God.

7
00:01:42,311 --> 00:01:44,061
That's why he's in trouble.

8
00:01:44,146 --> 00:01:45,313
George is a good guy.

9
00:01:46,482 --> 00:01:47,732
Give him a break, God.

10
00:01:47,816 --> 00:01:49,942
I love him, dear Lord.
"
```




# Lex the srt file into tokens

```{r}
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Define the regex for each token
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
srt_regexes <- c(
  time  = "\\d+:\\d+:\\d+,\\d+",
  link  = "\\s*-->\\s*",
  index = "^\\d+$",
  text  = "^.+?$"
)


#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Collapse the file into a single string
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
srt <- paste(enc2utf8(srt), collapse = "\n")

#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Split the file by regex, and drop the 'link' between times
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tokens <- lex(srt, multiline = TRUE, srt_regexes)
tokens <- tokens[names(tokens) != 'link']

tokens
```


# Parse raw tokens into a data.frame

```{r}
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Merge together runs of text
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
rl    <- unclass(rle(names(tokens)))
end   <- cumsum(rl$lengths)[rl$values == 'text']
len   <- rl$lengths[rl$values == 'text']
start <- end - len + 1

text <- mapply(function(start, end) {
  paste(tokens[start:end], collapse = "\n")
}, start, end)


#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Extract index and time vectors
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
indices <- as.integer(tokens[names(tokens) == 'index'])
times   <- tokens[names(tokens) == 'time']

#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Munge into data.frame
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
data.frame(
  index  = indices,
  start  = times[c(T, F)],
  end    = times[c(F, T)],
  text   = text,
  stringsAsFactors = FALSE
)
```