<tool id="icfree_learner" name="iCFree learner" version="@TOOL_VERSION@+galaxy@VERSION_SUFFIX@" license="MIT">
    <description>Active learning and model training</description>
    <macros>
        <import>macros.xml</import>
    </macros>
    <expand macro="requirements"/>
    <command detect_errors="exit_code"><![CDATA[
        #import os
        cp '$input_param_tsv' param.tsv &&
        unzip '$input_folder_zip' -d indir &&
        (find indir -type f | while read file; do mv "\$file" indir; done) &&
        python -m icfree.learner
            --data_folder indir
            --parameter_file param.tsv
            --output_folder 'outdir'
            #if str($adv.name_list) != ''
                --name_list '$adv.name_list'
            #end if
            #if str($adv.test) == 'true'
                --test
            #end if
            --nb_rep '$adv.nb_rep'
            #if str($adv.flatten) != ''
                --flatten
            #end if
            #if str($adv.seed_cond.seed_param) == 'not_random'
                --seed '$adv.seed_cond.seed'
            #end if
            --nb_new_data_predict '$adv.nb_new_data_predict'
            --nb_new_data '$adv.nb_new_data'
            --parameter_step '$adv.parameter_step'
            --n_group '$adv.n_group'
            --km '$adv.km'
            --ks '$adv.ks'
            --save_plot
            --verbose && ls 'outdir'
    ]]></command>
    <inputs>
        <param name="input_folder_zip" type="data" format="zip" label="Zip folder containing the data files" help="Zip folder containing the data file"/>
        <param name="input_param_tsv" type="data" format="tabular" label="Parameter values for the experiments" help="Parameter values for the experiment"/>
        <section name="adv" title="Advanced Options" expanded="false">
            <param name="name_list" type="text" value="" label="Labels of the feature list" help="A comma-separated string of column names or identifiers, converted to a list of strings representing columns that contain labels (y). This separates y columns from the rest (X features). (Default: Yield1,Yield2,Yield3,Yield4,Yield5)" />
            <param name="test" type="boolean" label="Validate the model" help="A flag for validating the model; not required to run inside the active learning loop. If not set, skip the validating step" checked="false" />
            <param name="nb_rep" type="integer" value="100" label="Number of test repetitions for validation the model behavior" help="The number of test repetitions for validating the model behavior. 80% of data is randomly separated for training, and 20% is used for testing." />
            <param name="flatten" type="boolean" label="Flattent feature data" help="A flag to indicate whether to flatten Y data. If set, treats each repetition in the same experiment independently; multiple same X values with different y outputs are modeled. Else, calculates the average of y across repetitions and only model with y average." />
            <param name="nb_new_data_predict" type="integer" value="1000" label="Number of new data points generated" help="The number of new data points sampled from all possible cases." />
            <param name="nb_new_data" type="integer" value="50" label="Number of new data points used" help="The number of new data points selected from the generated ones. These are the data points labeled after active learning loops. `nb_new_data_predict` must be greater than `nb_new_data` to be meaningful." />
            <param name="parameter_step" type="integer" value="10" label="Step size used to decrement the maximum predefined concentration sequentially" help="The step size used to decrement the maximum predefined concentration sequentially. For example, if the maximum concentration is `max`, the sequence of concentrations is calculated as: `max - 1 * parameter_step`, `max - 2 * parameter_step`, `max - 3 * parameter_step`, and so on. Each concentration is a candidate for experimental testing. Smaller steps result in more possible combinations to sample." />
            <param name="n_group" type="integer" value="15" label="Number of clusters" help="Parameter for the cluster margin algorithm, specifying the number of groups into which generated data will be clustered." />
            <param name="km" type="integer" value="50" label="Number of data points for the first selection" help="Parameter for the cluster margin algorithm, specifying the number of data points for the first selection. Ensure `nb_new_data_predict > ks > km`." />
            <param name="ks" type="integer" value="20" label="Number of data points for the second selection" help="Parameter for the cluster margin algorithm, specifying the number of data points for the second selection. This is also similar to `nb_new_data`." />
            <!-- Seed -->
            <conditional name="seed_cond">
                <param name="seed_param" type="select" label="Seed" help="Choose a seed or let it as random">
                    <option value="not_random" selected="true">fixed</option>
                    <option value="random">random</option>
                </param>
                <when value="random"/>
                <when value="not_random">
                    <param name="seed" type="text" value="85" label="Seed value" help="Only integer allowed">
                        <validator type="empty_field" message="Not empty, select random"/>
                        <validator type="regex" message="Only integer allowed">^(?:\d+)$</validator>
                    </param>
                </when>
            </conditional>
        </section>
    </inputs>
    <outputs>
        <collection name="output_csv" type="list" label="${tool.name} - Data">
            <discover_datasets pattern="(?P&lt;name&gt;.*).csv" format="csv" directory="outdir" />
        </collection>
        <collection name="output_png" type="list" label="${tool.name} - Plot">
            <discover_datasets pattern="(?P&lt;name&gt;.*).png" format="png" directory="outdir" />
        </collection>
    </outputs>
    <tests>
        <test>
            <!-- python -m icfree.learner -data_folder learner -parameter_file learner.input.param.tsv -output_folder tmp -save_plot -verbose -seed 85 -->
            <param name="input_folder_zip" value="learner.input.folder.zip" />
            <param name="input_param_tsv" value="learner.input.param.tsv" />
            <output_collection name="output_csv" type="list" count="1">
                <element name="next_sampling_ei50" ftype="csv" >
                    <assert_contents>
                        <has_n_lines n="51" />
                    </assert_contents>
                </element>
            </output_collection>
            <output_collection name="output_png" type="list" count="4">
                <element name="EI selected">
                    <assert_contents>
                        <has_size value="77k" delta="1k"/>
                    </assert_contents>
                </element>
                <element name="EI">
                    <assert_contents>
                        <has_size value="36k" delta="1k"/>
                    </assert_contents>
                </element>
                <element name="Train_Test">
                    <assert_contents>
                        <has_size value="64k" delta="1k"/>
                    </assert_contents>
                </element>
                <element name="Yield evolution through each active learning query">
                    <assert_contents>
                        <has_size value="19k" delta="1k"/>
                    </assert_contents>
                </element>
            </output_collection>
        </test>
        <test>
            <!-- python -m icfree.learner -data_folder learner -parameter_file learner.input.param.tsv -output_folder tmp2 -name_list "Yield1,Yield2" -nb_rep 5 -flatten -seed 85 -nb_new_data_predict 20 -nb_new_data 2 -parameter_step 2 -n_group 3 -km 5 -ks 2 -save_plot -verbose -->
            <param name="input_folder_zip" value="learner.input.folder.zip" />
            <param name="input_param_tsv" value="learner.input.param.tsv" />
            <param name="name_list" value="Yield1,Yield2" />
            <param name="nb_rep" value="5" />
            <param name="flatten" value="true" />
            <param name="nb_new_data_predict" value="20" />
            <param name="nb_new_data" value="2" />
            <param name="parameter_step" value="2" />
            <param name="n_group" value="3" />
            <param name="km" value="5" />
            <param name="ks" value="2" />
            <output_collection name="output_csv" type="list" count="1">
                <element name="next_sampling_ei5" file="learner.output.data.2.csv" ftype="csv" >
                    <assert_contents>
                        <has_n_lines n="6" />
                    </assert_contents>
                </element>
            </output_collection>
            <output_collection name="output_png" type="list" count="4">
                <element name="EI selected">
                    <assert_contents>
                        <has_size value="24k" delta="1k"/>
                    </assert_contents>
                </element>
                <element name="EI">
                    <assert_contents>
                        <has_size value="36k" delta="1k"/>
                    </assert_contents>
                </element>
                <element name="Train_Test">
                    <assert_contents>
                        <has_size value="60k" delta="1k"/>
                    </assert_contents>
                </element>
                <element name="Yield evolution through each active learning query">
                    <assert_contents>
                        <has_size value="19k" delta="1k"/>
                    </assert_contents>
                </element>
            </output_collection>
        </test>
    </tests>
    <help><![CDATA[
Learner
=======
Active learning and model training

]]></help>
    <expand macro="creator"/>
    <expand macro="citation"/>
</tool>