---
title: "ProTrackR vs ProTrackR2"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{ProTrackR vs ProTrackR2}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

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

## ProTracker2 and its predecessor

The predecessor to this package ([ProTrackR](https://pepijn-devries.github.io/ProTrackR/)) was
entirely programmed in R. Although technically possible, it was challenging and slow with recursive algorithms.
The new version is a complete overhaul in C/C++, based on Olav Sørensen's
[ProTracker clone](https://16-bits.org/pt2.php). With it, came some design changes which are worth mentioning
here.

## Overview of differences

The table below summarises the differences between ProTrackR2 and its predecessor.

```{r tab-dif, echo=FALSE, message=FALSE, warning=FALSE}
library(kableExtra)
load("../tests/testthat/_openmpt/testcasestate.Rdata")
testresult <- table(openmpt_testfiles$state)
```

| Feature                             | ProTrackR                         | ProTrackR2                         |
|-------------------------------------|-----------------------------------|------------------------------------|
| [Effect commands](#effect-commands) | Limited set implemented           | All PT2.3d effects implemented     |
| [Infrastructure](#infrastructure)   | R script                          | Compiled C/C++                     |
| [File readers](#file-readers)       | Optimized for format preservation | Optimized for PT2.3d compatibility |
| [Audio output](#audio-output)       | tuneR S4 Wave class               | audio s3 audioSample class         |
| OpenMPT test cases                  | Passes 6 out of 12 selected tests | Passes at least `r testresult[["Success"]]` of the `r sum(testresult)` tests |

## Effect commands

ProTracker uses specific codes to apply certain effects or position jumps. `ProTrackR` implements only
a subset of these effects, whereas `ProTrackR2` has implemented all ProTracker compatible effects.

The predecessor only partly implemented arpeggio and setting finetune. It did not implement
glissando, sample filtering (E8) and loop reversal (EF). All these effects are implemented in
the current package.

For a full overview of effect commands see `vignette("effect_commands")`.

## Infrastructure

By switching to `C` and `C++` compiled code, the new package gained a significant
performance boost (see [benchmark results](#benchmark)). Where the in the predecessor
the module was represented by a vector of `raw` data, an `externalptr` to a `C struct`
is used in the current package. This required a slightly different approach to handle
these objects. In order to avoid confusion about the syntax, it was completely redesigned
in the successor.

### Class type objects

ProTrackR uses [s4 class](https://adv-r.hadley.nz/s4.html) objects. It has elegant semantics
and due to its strict nature, has a lot of build-in safety mechanisms. But consequently,
they are not straightforward and easy to work with.

Therefore, in ProTrackR2 I chose to work with the [s3 class](https://adv-r.hadley.nz/s3.html)
objects. They are easier to work with and perhaps more intuitive to most R users.
If you are coming to ProTrackR2 from ProTrackR, here is an overview of equivalent classes:

| ProTrackR s4 class       | ProTrackR equivalent s3 class | Description and remarks                   |
|--------------------------|-------------------------------|-------------------------------------------|
| `?ProTrackR::PTModule`   | `pt2_new_mod()`               | Overarching class representing a module   |
| `ProTrackR::PTCell()`    | `pt2_cell()`                  | Smallest structure in a pattern table     |
| No explicit class        | `pt2_note()`                  | Code representing a specific note         |
| No explicit class        | `pt2_command()`               | Code representing sound effect or trigger |
| `ProTrackR::PTBlock()`<br>`ProTrackR::PTTrack()` | `pt2_celllist()` | ProTrackR distinguishes between blocks (matrices) and tracks (columns) of cells. ProTrackR2 only considers a list of cells |
| `ProTrackR::PTPattern()` | `pt2_pattern()`               | A pattern table                           |
| `ProTrackR::PTSample()`  | `pt2_sample()`                | An audio sample                           |

These different class styles also have consequences for your workflow. Below are some
examples of how to achieve similar operations using the different packages.

```{r diffs, message=FALSE}
library(ProTrackR)
library(ProTrackR2)

## *Old* way of creating a blank module:
mod_old <- new("PTModule")
## *New* way of creating a blank module:
mod_new <- pt2_new_mod("my mod")

## *Old* reading a module:
## Note that `ProTrackR` does not support compressed modules,
## so we first need to create a file that the old library can read:
file_decompressed <- tempfile()
pt2_write_mod(pt2_read_mod(pt2_demo()), file_decompressed)
## Now the old library can actually read it
mod_old <- read.module(file_decompressed)
## *New* way of creating a blank module:
mod_new <- pt2_read_mod(pt2_demo())

## *Old* way of getting a sample
samp_old <- PTSample(mod_old, 1L)
## *New* way of getting a sample
samp_new <- mod_new$samples[[1L]]

## *Old* way of getting a pattern
patt_old <- PTPattern(mod_old, 1L)
## *New* way of getting a pattern
patt_new <- mod_new$patterns[[1L]]

## *Old* way of getting a cell
cell_old <- PTCell(mod_old, 1L, 1L, 1L)
## *New* way of getting a cell
##
##   First cell in cell-list ---------------+
##   First column in pattern ----------+    |
##      First row in pattern ------+   |    |
##             First pattern -+    |   |    |
##                            |    |   |    |
##                            V    V   V    V
cell_new <- mod_new$patterns[[1L]][1L, 1L][[1L]]

print(cell_old)
print(cell_new)
```

### Benchmark

A benchmark test where the same module (the one provided with this package) is rendered with
both `ProTrackR` and `ProTrackR2`. The settings for both tests were similar and performed
on the same system and repeated 10 times. On average `ProTrackR2` renders 8.8 times faster
than `ProTrackR`.

## File readers

While reading ProTracker modules, the predecessor preserved the data in the file. It only
modified / fixed data when requested by the user. The current package will always sanitise
data while reading it, making it compatible with ProTracker 2.3d. The current reader is
also a bit more flexible and allows to read more exotic formats. It even allows you to
read files compressed with PowerPacker.

### Alternatives

If you want even more flexibility, check out the [openmpt package](https://pepijn-devries.github.io/openmpt/).
It uses [libopenmpt](https://lib.openmpt.org/) to read and play modules. This library
has a more extensive set of supported [file formats](https://wiki.openmpt.org/Manual:_Module_formats).
The downside is that it does not allow you to modify or save modules.

## Audio output

The predecessor used [tuneR](https://cran.r-project.org/package=tuneR) objects to store
rendered audio. In the current package we use [audio](https://cran.r-project.org/package=tuneR)
objects. This switch was made as the `S3` class objects from 'audio' are easier to handle than the
stricter and formal `S4` class objects from 'tuneR'. If you wish to use the advanced features
from the 'tuneR' package, this is still possible as both formats can be converted relatively easy.

```{r audio-tuner, eval=FALSE}
## Load demo module
mod <- pt2_read_mod(pt2_demo())

## render 'audioSample' object
mod_audio <- pt2_render(mod)

## Convert from 'audio::audioSample' to 'tuneR::Wave' object:
if (requireNamespace("tuneR")) {
  mod_tuneR <-
    tuneR::Wave(
      left      = as.integer(2^15*unclass(mod_audio[1,])),
      right     = as.integer(2^15*unclass(mod_audio[2,])),
      samp.rate = attributes(mod_audio)$rate,
      bit       = attributes(mod_audio)$bits
    )
}
```