---
title: "Creating new `ChromBackend` classes for Chromatograms"
output:
    BiocStyle::html_document:
        toc_float: true
vignette: >
    %\VignetteIndexEntry{Creating new `ChromBackend` class for Chromatograms}
    %\VignetteEngine{knitr::rmarkdown}
    %\VignetteEncoding{UTF-8}
    %\VignettePackage{Chromatograms}
    %\VignetteDepends{Chromatograms,BiocStyle,S4Vectors,IRanges}
---

```{r style, echo = FALSE, results = 'asis', message=FALSE}
BiocStyle::markdown()
```

**Package**: `r Biocpkg("Chromatograms")`<br />
**Authors**: `r packageDescription("Chromatograms")[["Author"]] `<br />
**Compiled**: `r date()`

```{r, echo = FALSE, message = FALSE}
library(Chromatograms)
library(BiocStyle)
```

# Introduction

Similar to the `r Biocpkg("Spectra")` package, the `r Biocpkg("Chromatograms")`
also separates the user-faced functionality to process and analyze
chromatographic mass spectrometry (MS) data from the code for storage and
*representation* of the data. The latter functionality is provided by
implementations of the `ChromBackend` class, further on called *backends*. This
vignette describes the `ChromBackend` class and illustrates on a simple example
how a backend extending this class could be implemented.

Contributions to this vignette (content or correction of typos) or requests for
additional details and information are highly welcome, ideally *via* pull
requests or *issues* on the package's [github repository](https://github.com/RforMassSpectrometry/Chromatograms).


# What is a `ChromBackend`?

The purpose of a backend class extending the virtual `ChromBackend` is to
provide the chromatographic MS data to the `Chromatograms` object, which is used
by the user to interact with - and analyze the data. The `ChromBackend` defines
the API that new backends need to provide so that they can be used with
`Chromatograms`. This API defines a set of methods to access the data. For many
functions default implementations exist and a dedicated implementation for a new
backend is only needed if necessary (e.g. if the data is stored in a way that a
different access to it would be better). In addition, a core set of variables
(data fields), the so called *core* chromatogram variables, is defined to
describe the chromatographic data. Each backend needs to provide these, but can
also define additional data fields. Before implementing a new backend it is
highly suggested to carefully read the following *Conventions and definitions*
section.


## Conventions and definitions

General conventions for chromatographic MS data of a `Chromatograms` are:

- One `Chromatograms` object is designed to contain multiple chromatographic
  data (not data from a single chromatogram).
- retention time values within each chromatogram are expected to be sorted
  increasingly.
- Missing values (`NA`) for retention time values are not supported.
- Properties (data fields) of a chromatogram are called *chromatogram
  variables*. While backends can define their own properties, a minimum required
  set of chromatogram variables **must** be provided by each backend (even if
  their values are empty). These *core chromatogram variables* are listed (along
  with their expected data type) by the `coreChromVariables()` function.
- `dataStorage` and `dataOrigin` are two special variables that define
  for each chromatogram where the data is (currently) stored and from where the
  data derived, respectively. Both are expected to be of
  type`character`. Missing values for `dataStorage` are not allowed.
- `ChromBackend` implementations can also represent purely *read-only* data
  resources. In this case only data accessor methods need to be implemented but
  not data replacement methods (i.e. `<-` methods that would allow to add or set
  variables. Read-only backends should implement the `isReadOnly()` method, that
  should then return `TRUE`. Note that backends for purely read-only resources
  could also implement a *caching* mechanism to (temporarily) store changes to
  the data locally within the object (and hence in memory). See information on
  the `MsBackendCached` in the `r Biocpkg("Spectra")` package for more details.


## Notes on parallel and chunk-wise processing

For parallel processing, `Chromatograms` splits the backend based on a defined
`factor` and processes each in parallel (or *in serial* if a `SerialParam` is
used). The splitting `factor` can be defined for `Chromatograms` by setting the
parameter `processingChunkSize`. Alternatively, through the
`backendParallelFactor()` method the backend can also *suggest* a `factor` that
should/could be used for splitting and parallel processing. The default
implementation for `backendParallelFactor()` is to return an empty `factor`
(`factor()`) hence not suggesting any preferred splitting.

Besides parallel processing, for on-disk backends (i.e., backends that don't
keep all of the data in memory), this chunk-wise processing can also reduce the
memory demand for operations, because only the peak data of the current chunk
needs to be realized in memory.


# API

The `ChromBackend` class defines core methods that have to be implemented by a
MS *backend* as well as *optional* methods for which a default implementation is
already available. These functions are described in sections *Required methods*
and *Optional methods*, respectively.

To create a new backend a class extending the virtual `ChromBackend` needs to be
implemented. In the following example we define a simple class that uses a
`data.frame` to store general properties (*chromatogram variables*) and a list of
`data.frame` for the retention time and intensity values of each chromatograms,
which represent the actual chromatographic MS data. These values are store in a 
`list`, where each element correspond to one chromatogram, as the 
number of values (*peaks*) can vary between chromatograms. We also provide a
basic constructor function that returns an empty instance of the new class.

```{r, message = FALSE}
library(Chromatograms)

#' Definition of the backend class extending ChromBackend
setClass("ChromBackendTest",
         contains = "ChromBackend",
         slots = c(
             chromData = "data.frame",
             peaksData = "list"
         ),
         prototype = prototype(
             chromData = data.frame(),
             peaksData = list()
         ))

#' Simple constructor function
ChromBackendTest <- function() {
    new("ChromBackendTest")
}
```

The 2 slots `@chromData` and `@peaksData` will be used to store the general
properties of the chromatograms and the actual chromatographic data, respectively.
each row in `chromData` will contain data for one chromatogram with the columns
being the different *chromatogram variables* (i.e. additional properties of a
chromatogram such as its m/z value or MS level) and each element in `@peaksData`
a `data.frame` with the retention time and intensity values representing thus
the *peaks* data of the respective chromatogram. This is only one of the possibly
many ways chromatographic data might be represented.

We should ideally also add some basic validity function that ensures the data to
be correct (valid). The function below simply checks that the number of rows of
the `@chromData` slot matches the length of the `@peaksData` slots.

```{r, message = FALSE}
#' Basic validation function
setValidity("ChromBackendTest", function(object) {
    if (length(object@peaksData) != nrow(object@chromData))
        return("length of 'peaksData' has to match the number of rows of ",
               "'chromData'")
    NULL
})
```

We can now create an instance of our new class with the `ChromBackendTest()`
function.

```{r}
#' Create an empty instance of ChromBackendTest
be <- ChromBackendTest()
be
```

A `show()` method would allow for a more convenient way how general information
of our object is displayed. Below we add an implementation of the `show()`
method.

```{r}
#' implementation of show for ChromBackendTest
setMethod("show", "ChromBackendTest", function(object) {
    cd <- object@chromData
    cat(class(object), "with", nrow(cd), "chromatograms\n")
})
be
```


## Required methods

Methods listed in this section **must** be implemented for a new class extending
`ChromBackend`. Methods should ideally also be implemented in the order they are
listed here. Also, it is strongly advised to write dedicated unit tests for each
newly implemented method or function already **during** the development.


### `dataStorage()`

The `dataStorage` chromatogram variable provides information how or where the
data is stored. The `dataStorage()` method should therefore return a `character`
vector of length equal to the number of chromatograms that are represented by
the object. The values for `dataStorage` can be any character value, except
`NA`. For our example backend we define a simple `dataStorage()` method that
simply returns the column `"dataStorage"` from the `@chromData` (as a
`character`).

```{r}
#' dataStorage method to provide information *where* data is stored
setMethod("dataStorage", "ChromBackendTest", function(object) {
    as.character(object@chromData$dataStorage)
})
```

Calling `dataStorage()` on our example backend will thus return an empty
`character` (since the object created above does not contain any data).

```{r}
dataStorage(be)
```


### `length()`

`length()` is expected to return an `integer` of length 1 with the total number
of chromatograms that are represented by the backend. For our example backend we
simply return the number of rows of the `data.frame` stored in the `@chromData`
slot.

```{r}
#' length to provide information on the number of chromatograms
setMethod("length", "ChromBackendTest", function(x) {
    nrow(x@chromData)
})
length(be)
```


### `backendInitialize()`

The `backendInitialize()` method should be called after creating an instance of
the backend class and is responsible for preparing (initializing) the backend
with data. This method can accept any parameters required by the backend to
load or initialize the data, such as file names, a database connection, or
objects containing the data. It is also recommended that the the special
chromatogram variables `dataStorage` and `dataOrigin` are set during 
`backendInitialize()`.

It is strongly recommended to validate the input data within the initialize
method. The advantage of performing these validity checks in
`backendInitialize()` rather than using `setValidity()` is that computationally
expensive operations/checks would only be performed once,during initialization, 
instead of each time values within the object are modified (e.g., through
subsetting or similar operations), which would occur with `setValidity()`. 

We also use the `validChromData()` and `validPeaksData()` functions to ensure
that core chromatogram variables and core peaks variables have the correct data
type. These checks verify that the`peaksData` contains only numeric values and
that the number of retention time and intensity values matches for each 
chromatogram.

Below we define a `backendInitialize()` method that accepts a `data.frame` 
containing chromatogram variables and a `list` with retention time and
intensity values for each chromatogram. 

```{r}
#' backendInitialize method to fill the backend with data.
setMethod(
    "backendInitialize", "ChromBackendTest",
    function(object, chromData, peaksData) {
        if (!is.data.frame(chromData))
            stop("'chromData' needs to be a 'data.frame' with the general",
                 "chromatogram variables")
        ## Defining dataStorage and dataOrigin, if not available
        if (is.null(chromData$dataStorage))
            chromData$dataStorage <- "<memory>"
        if (is.null(chromData$dataOrigin))
            chromData$dataOrigin <- "<user provided>"
        ## Validate the provided data
        validChromData(chromData)
        validPeaksData(peaksData)
        ## Fill the object with data
        object@chromData <- chromData
        object@peaksData <- peaksData
        object
    })
```

In addition to adding the data to object, the function also defined the
`dataStorage` and `dataOrigin` spectra variables. The purpose of these two
variables is to provide some information on where the data is currently stored
(*in memory* as in our example) and from where the data is originating.

We can now create an instance of our backend class and fill it with data. We
thus first define our MS data and pass this to the `backendInitialize()` method.

```{r}
# A data.frame with chromatogram variables.
cdata <- data.frame(msLevel = c(1L, 1L),
                    mz = c(112.2, 123.3))

# Retention time and intensity values for each chromatogram.
pdata <- list(
  data.frame(rtime = c(12.4, 12.8, 13.2, 14.6),
       intensity = c(123.3, 153.6, 2354.3, 243.4)),
  data.frame(rtime = c(45.1, 46.2),
       intensity = c(100, 80.1))
)

#' Create and initialize the backend
be <- backendInitialize(ChromBackendTest(),
                        chromData = cdata, peaksData = pdata)
be
```

This `backendInitialize()` implementation should assure data
validity and integrity. Below we use this function again to create our backend
instance.

The `backendInitialize()` method that we implemented for our backend class
expects the user to provide the full MS data. It would alternatively also be
possible to implement a method that takes data file names as input from which
the function can then import the data. The purpose of the `backendInitialize()`
method is to *initialize* and prepare the data in a way that it can be accessed
by a `Chromatograms` object. Whether the data is actually loaded into memory or
simply referenced and loaded upon request does not matter as long as the backend
is able to provide the data though its accessor methods when requested by the
`Chromatograms` object.


### `chromVariables()`

The `chromVariables()` method should return a `character` vector with the names
of all available chromatogram variables of the backend. While a backend class
should support defining and providing their own variables, each `ChromBackend`
class **must** provide also the *core chromatogram variables* (in the correct
data type). These can be listed by the `coreChromVariables()` function:

```{r}
#' List core chromatogram variables along with data types.
coreChromVariables()
```

A typical `chromVariables()` method for a `ChromBackend` class will thus be
implemented similarly to the one for our `ChromBackendTest` test backend: it
will return the names for all available chromatogram variables that can be 
called by `chromData()` within the backend object. There is a default
implementation for `chromVariables()` that will return the core chromatogram 
variables. However if a backend class defines additional chromatogram variables,
the `chromVariables()` method should be implemented to return the names of
these additional variables as well.

```{r}
#' Accessor for available chromatogram variables
setMethod("chromVariables", "ChromBackendTest", function(object) {
    union(names(object@chromData), names(coreChromVariables()))
})

chromVariables(be)
```

### `chromData()`

The `chromData` method should return the **full** chromatogram data within a
backend as a `data.frame` object. A parameter `columns` should allow to define
the names of the variables that should be returned. A parameter `drop` should
also be implemented to allow for the calling of one column while still 
controlling the return type.
Each row in this data frame should represent one chromatogram, each column a 
chromatogram variable. The `data.frame` **must** provide values (even if they
are `NA`) for **all** requested chromatogram variables of the backend
(**including** the core chromatogram variables). The `fillCoreChromVariables()` 
function from the *Chromatograms* package allows to *complete* (fill) a
provided `data.frame` with eventually missing core chromatogram variables:

```{r}
#' Get the data.frame with the available chrom variables
be@chromData
#' Complete this data.frame with missing core variables
fillCoreChromVariables(be@chromData)
```

We can thus use this function to add eventually missing core chromatogram
variables in the `chromData` implementation for our backend:

```{r}
#' function to extract the full chromData
setMethod(
    "chromData", "ChromBackendTest",
    function(object, columns = chromVariables(object),
             drop = FALSE) {
      if (!any(chromVariables(object) %in% columns))
        stop("Some of the requested Chromatogram variables are not ",
                   "available")
      res <- fillCoreChromVariables(object@chromData)
      res <- res[, columns, drop = drop]
      res
      })
```

We can now use `chromData()` to either extract the full chromatogram data from
the backend, or only the data for selected variables.

```{r}
#' Extract the full data
chromData(be)

#' Selected variables
chromData(be, c("mz", "msLevel"))

#' Only missing core spectra variables
chromData(be, c("collisionEnergy", "mzMin"))
```

### `peaksVariables()`

The `peaksVariables()` function is supposed to provide the names of the
available *peaks variables*. If additional peaks variables would be available, 
these could also be listed by the `peaksVariables()` method.
There is a default implementation for `peaksVaraibles()` that will return the
core peaks variables. However if a backend class defines additional
peaks variables, the `peaksVariables()` method should be implemented to
return the names of these additional variables as well.

```{r}
setMethod("peaksVariables", "ChromBackendTest", function(object) {
  union(names(corePeaksVariables()), names(object@peaksData[[1]]))
})
```

We can now see what peaks variables are present in our object:

```{r}
peaksVariables(be)
```

### `peaksData()`

The `peaksData()` method extracts the chromatographic data (*peaks*), i.e., the
chromatograms' retention time and intensity values. This data is returned as a
`list` of `data.frame`, with one array per chromatogram with columns being the
*peaks variables* (retention time and intensity values) and rows the individual
data pairs. Each backend must provide retention times and intensity values with
this method, but additional peaks variables (columns) are also supported.

In a similar way as for the chromatogram variables, a backend should support 
defining and providing their own variables and each `ChromBackend`
class **must** provide also the *core peaks variables* (in the correct
data type). These can be listed by the `corePeaksVariables()` function:

```{r}
corePeaksVariables()
```

Below we implement the `peaksData()` method for our backend.

```{r}
#' method to extract the full chromatographic data as list of arrays
setMethod(
    "peaksData", "ChromBackendTest",
    function(object, columns = peaksVariables(object), drop = FALSE) { 
      if (!all(columns %in% peaksVariables(object)))
        stop("Some of the requested peaks variables are not available")
      res <- lapply(object@peaksData, function(x) x[, columns, drop = drop])
      res
    })
```

And with this method we can now extract the peaks data from our backend.

```{r}
#' Extract the *peaks* data (i.e. intensity and retention times)
peaksData(be)
```

Since the `peaksData()` method is the main function used by a `Chromatograms` to
retrieve data from the backend (and further process the values), this method
should be implemented in an efficient way.


### `[`

The `[` method allows to subset `ChromBackend` objects. This operation is
expected to reduce a `ChromBackend` object to the selected chromatograms without
changing values for the subset chromatograms. The method should support to
subset by indices or logical vectors and should also support duplicating
elements (i.e., when duplicated indices are used) as well as to subset in
arbitrary order. An error should be thrown if indices are out of bounds, but the
method should also support returning an empty backend with `[integer()]`. The
`MsCoreUtils::i2index` function can be used to check and convert the provided
parameter `i` (defining the subset) to an integer vector.

Below we implement a possible `[` for our test backend class. We ignore the
parameters `j` from the definition of the `[` generic, since we treat our data
to be one-dimensional (with each chromatogram being one element).

```{r}
#' Main subset method.
setMethod("[", "ChromBackendTest", function(x, i, j, ..., drop = FALSE) {
  i <- MsCoreUtils::i2index(i, length = length(x))
  x@chromData <- x@chromData[i, ]
  x@peaksData <- x@peaksData[i]
  x
})
```

We can now subset our backend to the last two chromatograms.

```{r}
a <- be[1]
chromData(a)
```

Or extracting the second chromatogram multiple times.

```{r}
a <- be[c(1, 1, 1)]
chromData(a)
```


### `$`

The `$` method is expected to extract a single chromatogram or peaks variable 
from a backend. Parameter `name` should allow to name the variable to return. 
Each `ChromBackend` **must** support extracting the core chromatogram and core
peaks variables with this method (even if no data might be available for that
variable). In our example implementation below we make use of the `chromData()`
method, but more efficient implementations might be possible as well.
Also, the `$` method should check if the requested variable is available and
should throw an error otherwise.

```{r}
#' Access a single chromatogram variable
setMethod("$", "ChromBackendTest", function(x, name) {
  if (name %in% union(chromVariables(x), names(coreChromVariables())))
    res <- chromData(x, columns = name, drop = TRUE)
  else if (name %in% peaksVariables(x))
    res <- peaksData(x, columns = name, drop = TRUE)
  else stop("The requested variable '", name, "' is not available")
  res
})
```

With this we can now extract the MS levels

```{r}
be$msLevel
```

or a core chromatogram variable without values in our example backend.

```{r}
be$precursorMz
```

or also the intensity values

```{r}
be$intensity
```

### `backendMerge()`

The `backendMerge()` method merges (combines) `ChromBackend` objects (of the
same type!) into a single instance. For our test backend we thus need to combine
the values in the `@chromData`, `@peaksData` slots. To support also
merging of `data.frame`s with different sets of columns we use the
`MsCoreUtils::rbindFill` function instead of a simple `rbind` (this function
joins data frames making an union of all available columns filling eventually
missing columns with `NA`).

```{r}
#' Method allowing to join (concatenate) backends
setMethod("backendMerge", "ChromBackendTest", function(object, ...) {
    res <- object
    object <- unname(c(list(object), list(...)))
    res@peaksData <- do.call(c, lapply(object, function(z) z@peaksData))
    res@chromData <- do.call(MsCoreUtils::rbindFill,
                             lapply(object, function(z) z@chromData))
    validObject(res)
    res
})
```

Testing the function by merging the example backend instance with itself.

```{r}
a <- backendMerge(be, be[2], be)
a
```


## Data replacement methods

As stated in the general description, `ChromBackend` implementations can also be
purely *read-only* resources allowing to just access, but not to replace
data. For these backends `isReadOnly()` should return `FALSE`. Data replacement
methods listed in this section would not need to be implemented. Our example
backend stores the full data in memory, within the object, and hence we can
easily change and replace values.

Since we support replacing values we also implement the `isReadOnly()` method
for our example implementation to return `FALSE` (instead of the default
`TRUE`).

```{r}
#' Default for backends:
isReadOnly(be)
```

```{r}
#' Implementation of isReadOnly for ChromBackendTest
setMethod("isReadOnly", "ChromBackendTest", function(object) FALSE)
isReadOnly(be)
```

All data replacement function are expected to return an instance of the same
backend class that was used as input.


### `chromData<-`

The main replacement method is `chromData<-` which should allow to replace the
chormtaogram variables content of a backend with new data. This data is 
expected to be provided as a `data.frame` (similar to the one returned by
`chromData()`). While values can be replaced, the number of chromatograms
before and after a call to `chromData<-` has to be the same. 

```{r}
#' Replacement method for the full chromatogram data
setReplaceMethod("chromData", "ChromBackendTest", function(object, value) {
  if (is(value, "DataFrame"))
    value <- as(value, "data.frame")
  if (!inherits(value, "data.frame"))
    stop("'value' is expected to be a 'data.frame'")
  if (length(object) && length(object) != nrow(value))
    stop("'value' has to be a 'data.frame' with ", length(object), " rows")
  validChromData(value)
  object@chromData <- value
  object
})
```

To test this new method we extract the full chromatogram data from our example
data set, add an additional column (chromatogram variable) and use `chromData<-`
to replace the data of the backend.

```{r}
d <- chromData(be)
d$new_col <- c("a", "b")

chromData(be) <- d
```

Check that we have now also the new column available.

```{r}
be$new_col
```


### `$<-`

The `$<-` method should allow to replace values for an existing chromatogram
variable or to add an additional variable to the backend. As with all
replacement methods, the `length` of `value` has to match the number of
chromatograms represented by the backend. For replacement of retention time or
intensity values we need also to ensure that the data would be correct after the
operation, i.e., that the number of retention time and intensity values per
chromatogram are the identical and that all retention time and intensity values
are numeric. Finally, we use the `validChromData()` function to ensure that,
after replacement, all core chromatogram variables have the correct data type.

```{r}
#' Replace or add a single chromatogram variable.
setReplaceMethod("$", "ChromBackendTest", function(x, name, value) {
  if (length(x) && length(value) != length(x))
    stop("length of 'value' needs to match the number of chromatograms ",
         "in object.")
  if (name %in% peaksVariables(x)) {
    if (!is.list(value))
      stop("The value for peaksData should be a list")
    for (i in seq_along(value)) {
      x@peaksData[[i]][[name]] <- value[[i]]
      validPeaksData(x@peaksData)
    }
  } else {
    x@chromData[, name] <- value
    validChromData(x@chromData)
  }
  x
})
```

We can thus replace an existing chromatogram variable, such as `msLevel`:

```{r}
#' Values before replacement
be$msLevel

#' Replace MS levels
be$msLevel <- c(3L, 2L)

#' Values after replacement
be$msLevel
```

We can also add a new chromatogram variables:

```{r}
#' Add a new chromatogram variable
be$name <- c("A", "B")
be$name
```

Or also replace intensity values. Below we replace the intensity values by
adding a value of +3 to each.

```{r}
#' Replace intensity values
be$msLevel3 <- be$msLevel + 3
be$msLevel3
```


### `peaksData<-`

The `peaksData<-` method should allow to replace the full peaks data (retention
time and intensity value pairs) of all chromatograms in a backend. As `value`, a
`list` of `data.frame` should be provided with columns names `"rtime"` and
`"intensity"`. Because the full peaks data is provided at once, this method can
(and should) support changing also the number of peaks per chromatogram (while
the methods like `rtime<-` or `$rtime` would not allow).

```{r}
#' replacement method for peaks data
setReplaceMethod("peaksData", "ChromBackendTest", function(object, value) {
  if (!is.list(value))
    stop("'value' is expected to be a list")
  if (length(object) && length(object) != length(value))
    stop("'value' has to be a list with ", length(object), " elements")
  validPeaksData(value)
  object@peaksData <- value
  object
})
```

With this method we can now replace the peaks data of a backend:

```{r}
#' Create a list with peaks matrices; our backend has 3 chromatograms
#' thus our `list` has to be of length 3
tmp <- list(
    data.frame(rtime = c(12.3, 14.4, 15.4, 16.4),
          intensity = c(200, 312, 354.1, 232)),
    data.frame(rtime = c(14.4),
          intensity = c(13.4))
)

be_2 <- be
#' Assign this peaks data to one of our test backends
peaksData(be_2) <- tmp

#' Evaluate that we properly added the peaks data
peaksData(be_2)
```


## Methods with available default implementations

Default implementations for the `ChromBackend` class are available for a large
number of methods. Thus, any backend extending this class will automatically
inherit these default implementations. Alternative, class-specific, versions
can, but don't need to be developed. The default versions are defined in the
*R/ChromBackend.R* file, and also listed in this section. If alternative
versions are implemented it should be ensured that the expected data type is
always used for core chromatogram variables. Use `coreChromVariables()` and 
`corePeaksVariables()` to list these mandatory data types.


### `backendParallelFactor()`

The `backendParallelFactor()` function allows a backend to suggest a preferred
way it could be split for parallel processing. The default implementation
returns `factor()` (i.e. a `factor` of length 0) hence not suggesting any
specific splitting setup.

```{r, eval = FALSE}
#' Is there a specific way how the object could be best split for
#' parallel processing?
setMethod("backendParallelFactor", "ChromBackend", function(object, ...) {
    factor()
})
```

```{r}
backendParallelFactor(be)
```

### `chromIndex()`

The `chromIndex()` function should return the value for the `"chromIndex"`
chromatogram variable. As a result, an `integer` of length equal to the number
of chromatograms in `object` needs to be returned. The default implementation
is:

```{r, eval = FALSE}
#' get the values for the chromIndex chromatogram variable
setMethod("chromIndex", "ChromBackend",
          function(object, columns = chromVariables(object)) {
              chromData(object, columns = "chromIndex", drop = TRUE)
          })
```

The result of calling this method on our test backend:

```{r}
chromIndex(be)
```


### `collisionEnergy()`

The `collisionEnergy()` function should return the value for the
`"collisionEnergy"` chromatogram variable. As a result, a `numeric` of length
equal to the number of chromatograms has to be returned. The default
implementation is:

```{r, eval = FALSE}
#' get the values for the collisionEnergy chromatogram variable
setMethod("collisionEnergy", "ChromBackend", function(object) {
    chromData(object, columns = "collisionEnergy", drop = TRUE)
})
```

The result of calling this method on our test backend:

```{r}
collisionEnergy(be)
```

The default replacement method for the `collisionEnergy` chromatogram variable
is:

```{r, eval = FALSE}
#' Default replacement method for collisionEnergy
setReplaceMethod(
    "collisionEnergy", "ChromBackend", function(object, value) {
    object$collisionEnergy <- value
    object
    })
```

This method thus makes use of the `$<-` replacement method we implemented
above. To test this function we replace the collision energy below.

```{r}
#' Replace the collision energy
collisionEnergy(be) <- c(20, 30)
collisionEnergy(be)
```


### `dataOrigin()`, `dataOrigin<-`

The `dataOrigin()` and `dataOrigin<-` methods return or set the value(s) for the
`"dataOrigin"` chromatogram variable. The values for this chromatogram variable
need to be of type `character` (the length equal to the number of
chromatograms). The default implementation for `dataOrigin()` is:

```{r, eval = FALSE}
#' Default implementation to access dataOrigin
setMethod("dataOrigin", "ChromBackend", function(object) {
    chromData(object, columns = "dataOrigin", drop = TRUE)
})
```

Below we use this method to access the values of the `dataOrigin` chromatogram
variable.

```{r}
#' Access the dataOrigin values
dataOrigin(be)
```

The default implementation for `dataOrigin<-` uses, like all defaults for
replacement methods, the `$<-` method:

```{r}
#' Default implementation of the `dataOrigin<-` replacement method
setReplaceMethod("dataOrigin", "ChromBackend", function(object, value) {
  object$dataOrigin <- value
  object
})
```

For our backend we can change the values of the `dataOrigin` variable:

```{r}
#' Replace the backend's dataOrigin values
dataOrigin(be) <- rep("from somewhere", 2)
dataOrigin(be)
```


### `dataStorage()`, `dataStorage<-`

Similarly, the `dataStorage()` and `dataStorage<-` methods should allow to get
or set the data storage chromatogram variable. Values of the `dataStorage`
chromatogram variable are expected to be of type `character` and for each
chromatogram in a backend one value needs to be defined (which can not be
`NA_character`). The default implementation for `dataStorage()` uses, like most
access methods, the `chromData()` function:

```{r, eval = FALSE}
#' Default implementation to access dataStorage
setMethod("dataStorage", "ChromBackend", function(object) {
    chromData(object, columns = "dataStorage", drop = TRUE)
})
```

Below we use this method to access the values of the `dataStorage` chromatogram
variable.

```{r}
#' Access the dataStorage values
dataStorage(be)
```

Note that this variable is supposed to provide information on the location where
the data is stored and hence for some type of backends it might not be possible
or advised to let the user change its values. For such backends a
`dataStorage<-` replacement method should be implemented specifically that
throws an error if values are replaced with eventually invalid values. The
default implementation for this method uses, like all defaults for replacement
methods, the `$<-` method:

```{r}
#' Default implementation of the `dataStorage<-` replacement method
setReplaceMethod("dataStorage", "ChromBackend", function(object, value) {
  object$dataStorage <- value
  object
})
```

For our backend we can change the values of the `dataStorage` variable:

```{r}
#' Replace the backend's datastorage values
dataStorage(be) <- c("here", "here")
dataStorage(be)
```


### `intensity()`, `intensity<-`

The `intensity()` and `intensity<-` methods allow to extract or set the
intensity values of the individual chromatograms represented by the backend. The
default for the `intensity()` function, which is expected to return a `list` of
`numeric` values with the intensity values of each chromatogram, uses the `peaksData()` method:

```{r, eval = FALSE}
#' Default method to extract intensity values
setMethod("intensity", "ChromBackend", function(object) {
    if (length(object)) {
        peaksData(object, column = "intensity", drop = TRUE)
    } else list()
})
```

The default replacement method for intensity values uses the `$<-` method:

```{r, eval = FALSE}
#' Default implementation of the replacement method for intensity values
setReplaceMethod("intensity", "ChromBackend", function(object, value) {
    pd <- peaksData(object)
    if (!is.list(value) || length(pd) != length(value))
        stop("'value' should be a list of the same length as 'object'")
    for (i in seq_along(pd)) {
        if (length(value[[i]]) != nrow(pd[[i]])) {
            stop(paste0("Length of 'value[[", i, "]]' does not match ",
                       "the number of rows in the intensity of chromatogram: ",
                       i, "'"))
        }
    }
    peaksData(object) <- lapply(seq_along(pd), function(i) {
        pd[[i]]$intensity <- value[[i]]
        return(pd[[i]])
    })
    object
})

```

```{r}
#' Replace intensity values
intensity(be)[[1]] <- intensity(be)[[1]] + 10
intensity(be)
```


### `isEmpty()`

The `isEmpty()` is a simple helper function to evaluate whether chromatograms
are *empty*, i.e. have no peaks (retention time and intensity values). It should
return a logical vector of length equal to the number of chromatograms in the
backend with `TRUE` if a chromatogram is empty and `FALSE` otherwise. The
default implementation uses the `lengths()` method (defined further below) that
returns for each chromatogram the number of available data points (peaks).

```{r, eval = FALSE}
#' Default implementation for `isEmpty()`
setMethod("isEmpty", "ChromBackend", function(x) {
    lengths(x) == 0L
})
```

```{r}
isEmpty(be)
```


### `isReadOnly()`

As discussed above, backends can also be *read-only*, hence only allowing to
access, but not to change any values (e.g. if the data is stored in a data base
and the connection to this data base does not support updating or replacing
data). In such cases, the default `isReadOnly()` method can be used, which
returns always `TRUE`:

```{r, eval = FALSE}
#' Default implementation of `isReadOnly()`
setMethod("isReadOnly", "ChromBackend", function(object) {
    TRUE
})

```

Backends that support changing data values should implement their own version
(like we did above) to return `FALSE` instead:

```{r}
isReadOnly(be)
```


### `length()`

The `length()` method should return a single `integer` with the total number of
chromatograms available through the backend. The default implementation for this
function is:

```{r, eval = FALSE}
#' Default implementation for `length()`
setMethod("length", "ChromBackend", function(x) {
    nrow(chromData(x, columns = "dataStorage"))
})
```

```{r}
length(be)
```


### `lengths()`

The `lengths()` function should return the number of data pairs (peaks;
retention time or intensity values) per chromatogram. The result should be an
`integer` vector (of length equal to the number of chromatograms in the backend)
with these counts. The default implementation uses the `intensity()` function.

```{r, eval = FALSE}
#' Default implementation for `lengths()`
setMethod("lengths", "ChromBackend", function(x) {
    lengths(intensity(x))
})
```

The number of peaks for our test backend:

```{r}
lengths(be)
```


### `msLevel()`, `msLevel<-`

The `msLevel()` and `msLevel<-` methods should allow extracting and setting the
MS level for the individual chromatograms. MS levels are encoded as `integer`,
thus, `msLevel()` must return an `integer` vector of length equal to the number
of chromatograms of the backend and `msLevel<-` should take/accept such a vector
as input. The default implementations for both methods are shown below.

```{r, eval = FALSE}
#' Default methods to get or set MS levels
setMethod("msLevel", "ChromBackend", function(object) {
    chromData(object, columns = "msLevel", drop = TRUE)
})
setReplaceMethod("msLevel", "ChromBackend", function(object, value) {
    object$msLevel <- value
    object
})
```

To test these we below replace the MS levels for our test data set and extract
these values again.

```{r}
msLevel(be) <- c(1L, 2L)
msLevel(be)
```


### `mz()`, `mz<-`

The `mz()` and `mz<-` methods should allow to extract or set the m/z value for
each chromatogram. The m/z value of a chromatogram is encoded as `numeric`,
thus, the methods are expected to return or accept a `numeric` vector of length
equal to the number of chromatograms. The default implementations are shown
below.

```{r, eval = FALSE}
#' Default implementations to get or set m/z value(s)
setMethod("mz", "ChromBackend", function(object) {
    chromData(object, columns = "mz", drop = TRUE)
})
setReplaceMethod("mz", "ChromBackend", function(object, value) {
    object$mz <- value
    object
})
```

We below set and extract these *target* m/z values.

```{r}
mz(be) <- c(314.3, 312.5)
mz(be)
```


### `mzMax()`, `mzMax<-`

The `mzMax()` and `mzMax<-` methods should allow to extract or set the upper m/z
boundary for each chromatogram. m/z values are encoded as `numeric`, thus, the
methods are expected to return or accept a `numeric` vector of length equal to
the number of chromatograms. The default implementations are shown below.

```{r, eval = FALSE}
#' Default implementations to get or set upper m/z limits
setMethod("mzMax", "ChromBackend", function(object) {
    chromData(object, columns = "mzMax", drop = TRUE)
})
setReplaceMethod("mzMax", "ChromBackend", function(object, value) {
    object$mzMax <- value
    object
})
```

Testing these functions by replacing the upper m/z boundary with new values.

```{r}
mzMax(be) <- mz(be) + 0.01
mzMax(be)
```


### `mzMin(), `mzMin<-`

The `mzMin()` and `mzMin<-` methods should allow to extract or set the lower m/z
boundary for each chromatogram. m/z values are encoded as `numeric`, thus, the
methods are expected to return or accept a `numeric` vector of length equal to
the number of chromatograms. The default implementations are shown below.

```{r, eval = FALSE}
#' Default methods to get or set the lower m/z boundary
setMethod("mzMin", "ChromBackend", function(object) {
  chromData(object, columns = "mzMin", drop = TRUE)
})

setReplaceMethod("mzMin", "ChromBackend", function(object, value) {
  object$mzMin <- value
  object
})
```

Testing these functions by replacing the lower m/z boundary with new values.

```{r}
mzMin(be) <- mz(be) - 0.01
mzMin(be)
```

### `precursorMz()`, `precursorMz<-`

The `precursorMz()` and `precursorMz<-` methods are expected to get or set the
values for the precursor m/z of each chromatogram (if available). These are
encoded as `numeric` (one value per chromatogram) - and if a value is not
available `NA_real_` should be returned. The default implementations are:

```{r, eval = FALSE}
#' Default implementations to get or set the precursorMz chrom variable
setMethod("precursorMz", "ChromBackend", function(object) {
    chromData(object, columns = "precursorMz", drop = TRUE)
})
setReplaceMethod("precursorMz", "ChromBackend", function(object, value) {
    object$precursorMz <- value
    object
})
```

Below we set and get the `precursorMz` chromatogram variable for our backend.

```{r}
precursorMz(be) <- c(NA_real_, 123.3)
precursorMz(be)
```


### `precursorMzMax()`, `precursorMzMax<-`

These methods are supposed to allow to get and set the `precursorMzMax`
chromatogram variable. The default implementations are:

```{r, eval = FALSE}
#' Default implementations for `precursorMzMax`
setMethod("precursorMzMax", "ChromBackend", function(object) {
    chromData(object, columns = "precursorMzMax", drop = FALSE)
})
setReplaceMethod("precursorMzMax", "ChromBackend", function(object, value) {
    object$precursorMzMax <- value
    object
})
```

Below we test these functions by setting and extracting the values for this
chromatogram variable.

```{r}
precursorMzMax(be) <- precursorMz(be) + 0.1
precursorMzMax(be)
```


### `precursorMzMin()`, `precursorMzMin<-`

These methods are supposed to allow to get and set the `precursorMzMin`
chromatogram variable. The default implementations are:

```{r, eval = FALSE}
#' Default implementations for `precursorMzMin`
setMethod("precursorMzMin", "ChromBackend", function(object) {
    chromData(object, columns = "precursorMzMin", drop = FALSE)
})
setReplaceMethod("precursorMzMin", "ChromBackend", function(object, value) {
    object$precursorMzMin <- value
    object
})
```

Below we test these functions by setting and extracting the values for this
chromatogram variable.

```{r}
precursorMzMin(be) <- precursorMz(be) - 0.1
precursorMzMin(be)
```


### `productMz()`, `productMz<-`

These methods are supposed to allow to get and set the `productMz` chromatogram
variable. The default implementations are:

```{r, eval = FALSE}
#' Default implementations for `productMz`
setMethod("productMz", "ChromBackend", function(object) {
    chromData(object, columns = "productMz", drop = TRUE)
})
setReplaceMethod("productMz", "ChromBackend", function(object, value) {
    object$productMz <- value
    object
})
```

Below we test these functions by setting and extracting the values for this
chromatogram variable.

```{r}
productMz(be) <- c(123.2, NA_real_)
productMz(be)
```


### `productMzMax()`, `productMzMax<-`

These methods are supposed to allow to get and set the `productMzMax`
chromatogram variable. The default implementations are:

```{r, eval = FALSE}
#' Default implementations for `productMzMax`
setMethod("productMzMax", "ChromBackend", function(object) {
    chromData(object, columns = "productMzMax", drop = FALSE)
})
setReplaceMethod("productMzMax", "ChromBackend", function(object, value) {
    object$productMzMax <- value
    object
})
```

Below we test these functions by setting and extracting the values for this
chromatogram variable.

```{r}
productMzMax(be) <- productMz(be) + 0.02
productMzMax(be)
```


### `productMzMin()`, `productMzMin<-`

These methods are supposed to allow to get and set the `productMzMin`
chromatogram variable. The default implementations are:

```{r, eval = FALSE}
#' Default implementations for `productMzMin`
setMethod("productMzMin", "ChromBackend", function(object) {
    chromData(object, columns = "productMzMin", drop = FALSE)
})
setReplaceMethod("productMzMin", "ChromBackend", function(object, value) {
    object$productMzMin <- value
    object
})
```

Below we test these functions by setting and extracting the values for this
chromatogram variable.

```{r}
productMzMin(be) <- productMz(be) - 0.2
productMzMin(be)
```


### `rtime()`, `rtime<-`

The `rtime()` and `rtime<-` methods allow to get and set the retention times of
the individual chromatograms of the backend. Similar to the method for the
intensity values described above they should return or accept a `NumericList`,
each element being a `numeric` vector with the retention time values of one
chromatogram. The default implementations of these methods are shown below.

```{r, eval = FALSE}
#' Default methods for `rtime()` and `rtime<-`
setMethod("rtime", "ChromBackend", function(object) {    
  if (length(object)) {
      peaksData(object, column = "rtime", drop = TRUE)
  } else list()
})

setReplaceMethod("rtime", "ChromBackend", function(object, value) {
    pd <- peaksData(object)
    if (!is.list(value) || length(pd) != length(value))
        stop("'value' should be a list of the same length as 'object'")
    for (i in seq_along(pd)) {
        if (length(value[[i]]) != nrow(pd[[i]])) {
            stop(paste0("Length of 'value[[", i, "]]' does not match ",
            "the number of rows in 'the rtime of chromatogram: ", i, "'"))
        }
    }
    peaksData(object) <- lapply(seq_along(pd), function(i) {
        pd[[i]]$rtime <- value[[i]]
        return(pd[[i]])
    })
    object
})
```

We below test this implementation replacing the retention times of our example
backend by shifting all values by 2 seconds.

```{r}
rtime(be)[[1]] <- rtime(be)[[1]] + 2
rtime(be)
```


### `split()`

The `split()` method should split the backend into a `list` of backends
containing subsets of the original backend. The default implementation uses the
default implementation of `split()` from R and should work in most cases. This
function uses the `[` method to subset/split the object.

```{r, eval = FALSE}
#' Default method to split a backend
setMethod("split", "ChromBackend", function(x, f, drop = FALSE, ...) {
    split.default(x, f, drop = drop, ...)
})
```

We below test this by splitting the backend into two subsets.

```{r}
split(be, f = c(1, 2, 1))
```


# Session information

```{r si}
sessionInfo()
```

# References