configure

#!/bin/bash

echo "Configuring the Unipept backend program."

erro() {
    echo "$@" >&2
}

sedscript="$(mktemp tmp.XXXXXXX.sed)"
trap "rm '$sedscript'" EXIT

# --------------------------------------------------------------------
# checking for dependencies


checkdep() {
    which $1 > /dev/null 2>&1 || hash $1 > /dev/null 2>&1 || {
        erro "Unipept backend requires ${2:-$1} to be installed."
        exit 1
    }
}

checkdep diff
checkdep cat
checkdep curl
checkdep grep
checkdep java
checkdep make
checkdep mkdir
checkdep mvn "Maven"
checkdep rm
checkdep tee
checkdep tr
checkdep uniq
checkdep wc
checkdep wget
checkdep xargs
checkdep umgap "umgap crate (for umgap buildindex)"

# --------------------------------------------------------------------
# Prompt functions

ask() {
    local answer
    erro -n "$1 [Y/n] "
    read answer
    while true; do
        case "${answer:-Yes}" in
            N*|n*) return 1;;
            Y*|y*) return 0;;
        esac
        erro -n "Please answer with [y]es or [n]o. "
        read answer
    done
}

confirm() {
    local question="$1" default="$2" answer=
    erro -n "$question [$default] "
    read answer
    echo "${answer:-$default}"
}

# --------------------------------------------------------------------
# Backing up the previous files

backup() {
    if [ -e "$1" ]; then
        erro "Seems like '$1' already exists."
        if ask "Should I make a backup?"; then
            i=1
            while [ -e "${1}.$i" ]; do i=$((i + 1)); done
            cp "$1" "${1}.$i"
            erro "Moved '$1' to '${1}.${i}'"
        fi
    fi
}

backup "latest-config"

# --------------------------------------------------------------------
# Reading and remembering defaults

if [ -f "latest-config" ]; then
    source "latest-config"
fi

cat > "latest-config" <<"HERE"
# Run `./configure` to configure the unipept backend makefile.
#
# Editing this file does not configure the unipept backend run, it will
# only change the suggested configuration when you run `./configure`.
HERE

option() {
    local file="latest-config" option="$1" default="$2" question="$3" value=
    value="$(confirm "$question" "${!option:-$default}")"
    echo "$option=\"$value\"" >> "$file"
    echo "s|<<<${option#CONFIG_}>>>|${value//|/\\|}|g" >> "$sedscript"
}

option "CONFIG_PEPTIDE_MIN_LENGTH" "5" "What is the minimum length (inclusive) for tryptic peptides?"
option "CONFIG_PEPTIDE_MAX_LENGTH" "50" "What is the maximum length (inclusive) for tryptic peptides?"
option "CONFIG_KMER_LENGTH" "9" "What is the length (k) of the K-mer peptides?"
option "CONFIG_TABDIR" "./data/tables" "Where should I store the final TSV files (large, single-write)?"
option "CONFIG_INTDIR" "./data/intermediate" "Where should I store intermediate TSV files (large, single-write, multiple-read?"
option "CONFIG_TAXDIR" "./data/taxon" "Where should I store and extract the downloaded taxon zip (small, single-write, single-read)?"
option "CONFIG_SRCDIR" "./data/sources" "Where should I store the downloaded source xml files (large, single-write, single-read)?"
option "CONFIG_JAVA_MEM" "6g" "How much memory should Java use?"
option "CONFIG_ENTREZ_BATCH_SIZE" "1000" "Which batch size should I use for communication with Entrez?"
option "CONFIG_CMD_SORT" "sort --buffer-size=80% --parallel=4" "Which sort command should I use?"
option "CONFIG_CMD_GZIP" "gzip -" "Which pipe compression command should I use?"
option "CONFIG_CMD_ZCAT" "zcat" "Which pipe decompression command (e.g. zcat, gzcat) should I use?"
option "CONFIG_CMD_UNZIP" "unzip -DD" "How do I unzip while discarding dates?"
option "CONFIG_CMD_SED" "sed" "What's my sed executable (e.g. sed, gsed)?"
option "CONFIG_CMD_AWK" "awk" "What's my gnu awk executable (e.g. awk, gawk)?"
option "CONFIG_CMD_MKTEMP" "mktemp" "What's my gnu mktemp executable (e.g. mktemp, gmktemp)?"
option "CONFIG_CMD_JOIN" "join" "What's my gnu join executable (e.g. join, gjoin)?"

sources_count="${#CONFIG_SOURCES[@]}"
if [ "$sources_count" == "0" ]; then
    CONFIG_SOURCES=('swissprot' 'http://ftp.ebi.ac.uk/pub/databases/uniprot/knowledgebase/uniprot_sprot.xml.gz'
                    'trembl' 'http://ftp.ebi.ac.uk/pub/databases/uniprot/knowledgebase/uniprot_trembl.xml.gz')
    sources_count=4
fi
echo "CONFIG_SOURCES=(" >> "latest-config"
for (( i = 0 ; i < sources_count ; i += 2 )); do
    if ask "Parse ${CONFIG_SOURCES[$i]} (${CONFIG_SOURCES[$((i + 1))]})?"; then
        echo "    '${CONFIG_SOURCES[$i]}' '${CONFIG_SOURCES[$((i + 1))]}'" >> "latest-config"
    fi
done
erro -n "Add another source by entering the name. An empty name cancels: "
read name
while [ -n "$name" ]; do
    erro -n "Where can I download this source (url)? "
    read url

    echo "    '$name' '$url'" >> "latest-config"

    erro -n "Add another source by entering the name. An empty name cancels: "
    read name
done
echo ")" >> "latest-config"

source "latest-config"
sources_count="${#CONFIG_SOURCES[@]}"

# --------------------------------------------------------------------
# None-configured variables

ENTREZ_URL="https://eutils.ncbi.nlm.nih.gov/entrez/eutils"
echo "s|<<<ENTREZ_URL>>>|$ENTREZ_URL|g" >> "$sedscript"

TAXON_URL="http://ftp.ncbi.nih.gov/pub/taxonomy/taxdmp.zip"
echo "s|<<<TAXON_URL>>>|$TAXON_URL|g" >> "$sedscript"

LOGADD='echo $$(date +"[%s (%F %T)]")'
echo "s|^\t<<<LOGADD>>>|\t@$LOGADD|" >> "$sedscript"
echo "s|<<<LOGADD>>>|echo|g" >> "$sedscript"

# --------------------------------------------------------------------
# Writing the configured files

backup_if_changed() {
    original="$1"
    tmpfile="$(mktemp tmp.XXXXXXXX)"
    cat "$2" > "$tmpfile"
    if [ -e "$original" ]; then
        if ! diff "$original" "$tmpfile" > /dev/null; then
            backup "$original"
        fi
    fi
    mv "$tmpfile" "$original"
}

source_files=""
source_inputs=""
for (( i = 0 ; i < sources_count ; i += 2 )); do
    name="${CONFIG_SOURCES[$i]}"
    url="${CONFIG_SOURCES[$((i + 1))]}"
    source_files="$source_files $CONFIG_SRCDIR/${name}.xml.gz"
    source_inputs="$source_inputs $name=<($CONFIG_CMD_ZCAT $CONFIG_SRCDIR/${name}.xml.gz)"
done
echo "s|<<<SOURCE_FILES>>>|$source_files|g" >> "$sedscript"
echo "s|<<<SOURCE_INPUTS>>>|$source_inputs|g" >> "$sedscript"

tmpmakefile="$(mktemp makefile.XXXXXXXX)"
$CONFIG_CMD_SED -f "$sedscript" makefile.in > "$tmpmakefile"

for (( i = 0 ; i < sources_count ; i += 2 )); do
    name="${CONFIG_SOURCES[$i]}"
    file="$CONFIG_SRCDIR/${name}.xml.gz"
    url="${CONFIG_SOURCES[$((i + 1))]}"
    cat >> "$tmpmakefile" <<HERE
$file:
	$LOGADD "Started downloading $name."
	mkdir -p $CONFIG_SRCDIR
	rm -f $file
	wget --progress=dot:giga "$url" -O $file
	$LOGADD "Finished downloading $name."

HERE
done

backup_if_changed "makefile"           "$tmpmakefile"
backup_if_changed "type_strains.sh"    <($CONFIG_CMD_SED -f "$sedscript" type_strains.sh.in)
backup_if_changed "createEcNumbers.sh" <($CONFIG_CMD_SED -f "$sedscript" createEcNumbers.sh.in)
backup_if_changed "createGoTerms.sh"   <($CONFIG_CMD_SED -f "$sedscript" createGoTerms.sh.in)
backup_if_changed "createInterProEntries.sh"   <($CONFIG_CMD_SED -f "$sedscript" createInterProEntries.sh.in)

rm "$tmpmakefile"
chmod a+x type_strains.sh
chmod a+x createEcNumbers.sh
chmod a+x createGoTerms.sh
chmod a+x createInterProEntries.sh