<tool id="interactive_tool_isee" tool_type="interactive" name="iSEE" version="1.0.0">
  <requirements>
    <container type="docker">biocontainers/isee-galaxy:v3.13_cv1.0.0</container>
  </requirements>

  <entry_points>
    <entry_point name="iSEE Single Cell Visualisation" requires_domain="True">
      <!-- This port is exposed by the docker container -->
      <port>8888</port>
    </entry_point>
  </entry_points>

  <command><![CDATA[

## Container wdir: /import

#set INPUT_PATH = "sce"
#set RSCRIPT = '/scripts/isee.R'

ln -s '$input.extra_files_path' '$INPUT_PATH' &&
ln -s '$isee_script' '$RSCRIPT' &&
cp '$isee_script' '$outfile' &&

## Redirect stderr - unfortunately necessary to stop R "warning" messages from
## inferring job status as "failed"
/scripts/run.sh '$RSCRIPT'

  ]]>
  </command>

  <configfiles>
    <configfile name="isee_script"><![CDATA[

## Stop warning messages being emitted from R while still allowing genuine job failure
## -----------------------------------------------------------------------------
devNull <- file("/dev/null", open = "wt")
sink(devNull, type = "message")


## Begin R script
## -----------------------------------------------------------------------------
library(iSEE)
library(HDF5Array)

sce_path <- 'sce'
sce <- loadHDF5SummarizedExperiment(sce_path)
sce <- registerAppOptions(sce, color.maxlevels=40)

categorical_color_fun <- function(n){
  if (n <= 37) {
    # Less than 37 colours, use something from colour brewer
    # (joining a bunch of palettes, best colours up front)
    multiset <-  c(
        RColorBrewer::brewer.pal(9, "Set1"),
        RColorBrewer::brewer.pal(8, "Set2"),
        RColorBrewer::brewer.pal(12, "Set3"),
        RColorBrewer::brewer.pal(8, "Dark2"))
    return(multiset[1:n])
  }
  else {
    # More that 37, well at least it looks pretty
    return(rainbow(n))
  }
}


ecm <- ExperimentColorMap(

  # The default is viridis::viridis
  # https://cran.r-project.org/web/packages/viridis/vignettes/intro-to-viridis.html#the-color-scales
  # Setting continous is entirely a matter of taste
  # Some find magma easier to read than viridis

  all_continuous = list(
    assays  = viridis::magma,
    colData = viridis::magma,
    rowData = viridis::magma
  ),
  all_discrete = list(
    colData = categorical_color_fun,
    rowData = categorical_color_fun
  )
)


# These options are all sce-contents agnostic.
initial_plots <- c(

  # Show umap with clusters by default
  ReducedDimensionPlot(
                   DataBoxOpen=TRUE,
                   ColorBy="Column data",
                   VisualBoxOpen=TRUE,
                   PanelWidth=6L),

  # Show gene expression plot separated (and coloured) by cluster, by default.
  FeatureAssayPlot(XAxis = "Column data",
                   DataBoxOpen=TRUE,
                   VisualBoxOpen=TRUE,
                   ColorBy="Column data",
                   PanelWidth=6L
                   ),
  # Gene list is better wide
  RowDataTable(PanelWidth=12L),

  # For cell level observations (QC.)
  ColumnDataPlot(PanelWidth=6L,
                 DataBoxOpen=TRUE,
                 VisualBoxOpen=TRUE )
)

app <- iSEE(sce,
            colormap=ecm,
            initial=initial_plots)

shiny::runApp(app, host="0.0.0.0", port=8888, quiet=TRUE, launch.browser=FALSE)

]]></configfile>
  </configfiles>

  <inputs>
    <param
      name="input"
      type="data"
      format="rdata.se"
      label="HDF5SummarizedExperiment: rdata.se (composite)"
      help="This datatype represents a HDF5SummarizedExperiment object as
        exported from R."
    />
  </inputs>

  <outputs>
    <data name="outfile" format="txt"
          label="${tool.name} on ${on_string}: Rscript" />
  </outputs>

  <tests>
    <test expect_num_outputs="1">
    </test>
  </tests>

  <help><![CDATA[

**Overview**

iSEE provides a web interface for visualising single-cell transcriptomic
datasets encapsulated as an HDF5SummarizedExperiment object.

Full documentation: https://isee.github.io/iSEE/

-----

**Input**

The tool takes a single HDF5SummarizedExperiment object as an input,
which can be uploaded with the composite datatype ``rdata.se``. This requires
two input files: ``se.rds`` and ``assays.h5``. Within R, an
``HDF5SummarizedExperiment`` object can be exported as a directory containing
these two input files with the function ``saveHDF5SummarizedExperiment()``
- this will create a folder containing the two files required above.
These utilities can be installed and imported within R from the ``HDF5Array``
package.

-----

**Useful links:**

- Documentation on ``HDF5SummarizedExperiment`` data object:

  https://rdrr.io/bioc/HDF5Array/man/saveHDF5SummarizedExperiment.html

- HDF5Array manual:

  https://bioc.ism.ac.jp/packages/3.7/bioc/manuals/HDF5Array/man/HDF5Array.pdf

]]></help>

  <citations>
    <citation type="bibtex">
      @article{rue2018isee,
        title={iSEE: interactive summarizedexperiment explorer},
        author={Rue-Albrecht, Kevin and Marini, Federico and Soneson,
          Charlotte and Lun, Aaron TL},
        journal={F1000Research},
        volume={7},
        year={2018},
        publisher={Faculty of 1000 Ltd}
      }
    </citation>
  </citations>
</tool>