shellfu/src/include-bash/sfdoc.sh

#!/bin/bash

shellfu import pretty

#
# Show hidden objects (modules, functions, variables)?
#
SFDOC_SHOW_HIDDEN=${SFDOC_SHOW_HIDDEN:-false}

sfdoc__export() {
    #
    # Export module doc as manpage
    #
    local format=$1             # format to export in
    local MName=$2              # name to use in headers
    local MPath=$3              # path to file to export
    local Encoding=${4:-utf8}   # file encoding
    local MVer                  # module version (if found)
    local exfun                 # export function
    local usage="usage: sfdoc__export FMT NAME PATH [ENCODING]"
    test -n "$format"   || { warn "$usage"; return 2; }
    test -n "$MPath"    || { warn "$usage"; return 2; }
    test -n "$MName"    || { warn "$usage"; return 2; }
    exfun="__sfdoc__export_as_$format"
    type -t "$exfun" >/dev/null || {
        warn "unknown format: $format"
        return 2
    }
    MVer="v$(shellfu _read_directive "module-version" "$MPath" 2>/dev/null)"
    test "$MVer" == "v" && MVer="(no version)"
    debug -v format MName MPath Encoding MVer
    "$exfun"
}

sfdoc__ls() {
    #
    # List all objects of type $1 in module file $2
    #
    local otype=$1
    local mpath=$2
    case $otype in
            f)
                grep -HE '^[[:alnum:]_]+\(\) \{' "$mpath" \
                  | sed -e 's|(.*||; s|\.sh:|.|; s|^.*/||'
                ;;
            v)
                grep -HE '^[[:alnum:]_]+=' "$mpath" \
                  | sed -e 's|=.*||; s|\.sh:|.|; s|^.*/||'
                ;;
            *)
                warn "bug: invalid object type: $otype"
                ;;
    esac \
      | __sfdoc__filter_hidden
}

sfdoc__ls_m() {
    #
    # List all module files
    #
    shellfu _list_mfiles \
      | sed 's|\.sh$||; s|.*/||;' \
      | __sfdoc__filter_hidden
}

__sfdoc__export_as_manpage() {
    #
    # Export module $Module doc as manpage
    #
    __sfdoc__export_as_pod \
      | pod2man \
            --name "${MName^^}" \
            --center "User Contributed Shellfu Documentation" \
            --section 3x \
            --date "$(date -I -r "$MPath")" \
            --release "$MName $MVer" \
            --utf8 -q \`
}

__sfdoc__export_as_markdown() {
    #
    # Export module $Module doc as Markdown
    #
    local variable
    local function
    local vheader_done=false
    local fheader_done=false
    debug -v module
    echo "$MName"
    echo "$MName" | tr -c '\n' '[=*]'
    echo
    __sfdoc__get_doc | __sfdoc__md_escapes
    echo
    for variable in $(sfdoc__ls v "$MPath" | cut -d. -f2);
    do
        debug -v variable
        $vheader_done || {
            echo
            echo Variables
            echo ---------
            vheader_done=true
        }
        echo
        echo "### \`\$$variable\` ###"
        echo ""
        __sfdoc__get_doc "v:$variable" | __sfdoc__md_escapes
        echo
    done
    for function in $(sfdoc__ls f "$MPath" | cut -d. -f2);
    do
        debug -v function
        $fheader_done || {
            echo
            echo Functions
            echo ---------
            fheader_done=true
        }
        echo
        echo "### \`$function()\` ###"
        echo ""
        __sfdoc__get_doc "f:$function" | __sfdoc__md_escapes
        echo
    done
}

__sfdoc__export_as_pod() {
    #
    # Export module $Module doc as POD
    #
    local variable
    local function
    local vheader_done=false
    local fheader_done=false
    local indented=false
    echo "true <<'=cut'"
    echo "=pod"
    echo
    echo "=head1 NAME"
    echo
    echo "$MName - $(__sfdoc__get_doc | head -1)"
    echo
    echo "=head1 DESCRIPTION"
    echo
    __sfdoc__get_doc
    echo
    for variable in $(sfdoc__ls v "$MPath" | cut -d. -f2);
    do
        debug -v variable
        $vheader_done || {
            echo
            echo "=head1 VARIABLES"
            vheader_done=true
            echo
            echo "=over 8"
            echo
            indented=true
        }
        echo
        echo "=item I<\$$variable>"
        echo
        __sfdoc__get_doc "v:$variable"
        echo
    done
    $indented && {
        echo "=back"
        echo
    }
    for function in $(sfdoc__ls f "$MPath" | cut -d. -f2);
    do
        debug -v function
        $fheader_done || {
            echo
            echo "=head1 FUNCTIONS"
            fheader_done=true
            echo
            echo "=over 8"
            echo
            indented=true
        }
        echo
        echo "=item I<$function()>"
        echo
        __sfdoc__get_doc "f:$function"
        echo
    done
    $indented && {
        echo "=back"
        echo
    }
    echo "=encoding $Encoding"
    echo "=cut"
}

__sfdoc__filter_body() {
    #
    # Filter part $1 of body on stdin
    #
    local part=$1
    case "$part" in
        f:*)  __sfdoc__filter_fun "${part:2}" ;;
        v:*)  __sfdoc__filter_var "${part:2}" ;;
        *) warn "bug: invalid part specification $part" ;;
    esac
}

__sfdoc__filter_doc() {
    #
    # Filter docstring for part $1 from object body on stdin
    #
    local part=$1
    case $part in
        "")  __sfdoc__filter_mdoc ;;
        v:*)  __sfdoc__filter_vdoc ;;
        f:*)  __sfdoc__filter_fdoc ;;
        *) warn "bug: invalid part specification: $part" ;;
    esac
}

__sfdoc__filter_fun() {
    #
    # From module body on stdin, filter out function named $1
    #
    # Function definition should look like:
    #
    #     foo_fun() {
    #         foo=$1
    #     }
    #
    # That is,
    #
    #  *  no `function` keyword,
    #  *  name starts the first line,
    #  *  name is followed by '() {' with no additional spaces,
    #  *  function end is denoted by line with a single '}'.
    #
    local fn_name=$1
    name=$fn_name perl -we '
        undef $/;
        my $name = $ENV{name};
        my ($fbody) = <> =~ m/^($name\(\) \{.*?^\}$)/ms;
        print "$fbody\n" if defined $fbody;
    '
}

__sfdoc__filter_fdoc() {
    #
    # Filter docstring from function body on stdin
    #
    # Look for:
    #
    #  1. line "^    #$" to open docstring - this must be
    #     the first one after function name (`fun() {`)
    #  2. block of consecutive docstring lines (i.e. matching
    #     "^    # " or "^    #$"
    #  3. last "^    #$" line to close docstring.
    #  4. Next line must not match any of the patterns above.
    #
    # For example, stdin like this:
    #
    #     myfun() {
    #         #
    #         # Do my thing with foo $1
    #         #
    #         # Detailed description, possibly spanning
    #         # several paragraph.
    #         #
    #         # The format here should be Markdown.
    #         #
    #         local foo=$1
    #         printf %p "${foo:4}"
    #     }
    #
    # would be read as:
    #
    #     Do my thing with foo $1
    #
    #     Detailed description, possibly spanning
    #     several paragraph.
    #
    #     The format here should be Markdown.
    #
    # However if we added following line right before the `local`
    # declaration
    #
    #         #FIXME: TODO/FIXME lines are not dropped properly
    #
    # it **will not** be considered part of the docstring. This is
    # to allow for special comments tools like TODO/FIXME or
    # data for lint-like tools that are not part of function
    # description.
    #
    perl -we '
        my $isdoc;
        while (<>) {
            if (m/^    #$/) {
                $isdoc = 1;
            } elsif (m/^    [^#]/) {
                exit;
            }
            next unless $isdoc;
            s/^    //;
            print;
        }' | __sfdoc__strip_doc
}

__sfdoc__filter_hidden() {
    #
    # From objects on stdin, print only visible unless $SFDOC_SHOW_HIDDEN
    #
    $SFDOC_SHOW_HIDDEN && cat && return
    grep -v ^_ | grep -v \\._
}

__sfdoc__filter_mdoc() {
    #
    # From module body on stdin, filter module doc
    #
    # Example:
    #
    #     #!/bin/bash
    #     #
    #     # legal stuff
    #
    #     shellfu import irrelevant_stuff
    #
    #     #
    #     # Module is cool this
    #     #
    #     # Ok so this is my
    #     # "module", heh...
    #     #
    #
    perl -we '
        sub ok_chunk {
            my $chunk = shift;
            my @lines = split "\n", $_;
            chomp @lines;
            return 0 unless (defined $lines[0] and $lines[0] eq "#");
            return 0 unless $lines[$#lines] eq "#";
            foreach (@lines) {
                return 0 unless (m/^#$/ or m/^# /);
            }
            return 1
        }
        my @chunks;
        my @lines;
        $/ = "\n\n";
        @chunks = <STDIN>;
        foreach (@chunks) {
            if (ok_chunk "$_") {
                s/\n+$//;
                print "$_\n";
                exit 0;
            }
        }
        exit 1
    ' | __sfdoc__strip_doc
}

__sfdoc__filter_var() {
    #
    # From module body on stdin, filter variable definition named $1
    #
    # Look for:
    #
    #  1. Empty line, followed by line with single hash sign
    #  2. Set of consecutive docstring lines (start with '# ')
    #  3. Another "#" followed by one or more variable definitions (start
    #     with $name or \w+, followed by equal sign), at least one of which
    #     must be $1
    #  4. Empty line or EOF
    #
    # and return what is found.
    #
    # For example:
    #
    #     #
    #     # My variable
    #     #
    #     my_var=foo
    #
    # is found if $1 is 'my_var'
    #
    # A more advanced example:
    #
    #     #
    #     # Bar variables
    #     #
    #     # These serve similar purpose; the difference is
    #     # obvious
    #     #
    #     bar_1=one
    #     bar_2=two
    #     bar_3=three
    #
    # is found if $1 is either 'bar_1', 'bar_2' or 'bar_3'.  This allows
    # for a docstring being shared among closely related variables.
    #
    local oname=$1    # object name
    local state       # void, edge, dstr, vstr, junk
    local o_end       # object ended; is complete
    local cache       # buffer of hope
    # parse and keep each var docstring in cache;
    # then decide whether or not print the cache
    cache=$(mktemp -t shellfu_doc.__sfdoc__filter_var.XXXXXXXX)
    state=junk
    o_end=false
    while read -r line;
    do
        case $state:$line in
            # looks like void
            void:)          state=void; o_end=false ;;
            edge:)          state=void; o_end=true  ;;
            dstr:)          state=void; o_end=true  ;;
            vstr:)          state=void; o_end=true  ;;
            junk:)          state=void; o_end=false ;;
            # looks like edge
            void:"#")       state=edge; o_end=false ;;
            edge:"#")       state=dstr; o_end=false ;;
            dstr:"#")       state=dstr; o_end=false ;;
            vstr:"#")       state=dstr; o_end=false ;;
            junk:"#")       state=junk; o_end=false ;;
            # looks like docstring
            void:"# "*)     state=junk; o_end=false ;;
            edge:"# "*)     state=dstr; o_end=false ;;
            dstr:"# "*)     state=dstr; o_end=false ;;
            vstr:"# "*)     state=dstr; o_end=false ;;
            junk:"# "*)     state=junk; o_end=false ;;
            # looks like junk
            void:"#"*)      state=junk; o_end=false ;;
            edge:"#"*)      state=junk; o_end=false ;;
            dstr:"#"*)      state=junk; o_end=false ;;
            vstr:"#"*)      state=junk; o_end=false ;;
            junk:"#"*)      state=junk; o_end=false ;;
            # looks like variable string
            void:*=*)       state=vstr; o_end=false ;;
            edge:*=*)       state=junk; o_end=false ;;
            dstr:*=*)       state=vstr; o_end=false ;;
            vstr:*=*)       state=vstr; o_end=false ;;
            junk:*=*)       state=junk; o_end=false ;;
            # looks like something else
            *)              state=junk; o_end=false ;;
        esac
        case $state in
            edge|dstr|vstr) echo "$line" >> "$cache" ;;
        esac
        if $o_end;
        then
            if grep -q "^$oname=" "$cache";   # it's our wanted object
            then
                cat "$cache"
                rm "$cache"
                return 0
            fi
            rm "$cache"
        fi
    done
    rm "$cache"
    return 1
}

__sfdoc__filter_vdoc() {
    #
    # From variable definition body on stdin, filter doc
    #
    # Example:
    #
    #     #
    #     # Bar variables
    #     #
    #     # These serve similar purpose; the difference is
    #     # obvious
    #     #
    #     bar_1=one
    #     bar_2=two
    #     bar_3=three
    #
    grep '^#' | __sfdoc__strip_doc
}

__sfdoc__get_doc() {
    #
    # Show doc for part $2 of module $1
    #
    local part=$1
    __sfdoc__get_part "$part" | __sfdoc__filter_doc "$part"
}

__sfdoc__get_part() {
    #
    # Print part $1 (f|v:name) of module $MPath
    #
    # Part has format
    #
    #     TYPE:NAME
    #
    # where TYPE is 'v' for variables and f for functions, and
    # NAME is name of the part.
    #
    # If part is not specified, whole module file is printed.
    #
    local part=$1
    test -n "$part" || { cat "$MPath"; return $?; }
    <"$MPath" __sfdoc__filter_body "$part"
}

__sfdoc__md_escapes() {
    #
    # Do additional escapes for Markdown
    #
    # In docstrings, references to variables are often done
    # without backticks;  add these to prevent interpreting
    # underscores in variable names as cursive.
    #
    perl -ne '
        my @bigwords = ();
        my @newwords = ();
        if (m|^    |) {
            print;                              # probably code -- leave be
        } else {
            @bigwords = split " ";
            foreach (@bigwords) {
                if (m|`\w+\(\)`|) {
                    push @newwords, $_;              # function: escaped
                } elsif (m|(.*)\b(\w+)\(\)(.*)|) {
                    push @newwords, "$1`$2()`$3";    # function: needs escape
                } elsif (m|`\$\w+`|) {
                    push @newwords, $_;              # variable: escaped
                } elsif (m|(.*)\$(\w+)(.*)|) {
                    push @newwords, "$1`\$$2`$3";    # variable: needs escape
                } else {
                    push @newwords, $_;              # a normal word
                }
            }
            print join " ", @newwords;
            print "\n";
        }
    '
}

__sfdoc__strip_doc() {
    #
    # Strip doc out of Shellfu docstring
    #
    # From e.g.
    #
    #     #
    #     # foo
    #     #
    #     # bar
    #     # baz
    #     #
    #
    # where first and last line are mandatory and each line
    # must be either '#' or '# *', get:
    #
    #     foo
    #
    #     bar
    #     baz
    #
    # just as when you're parsing this text by eyes.
    #
    head -n -1 \
      | tail -n +2 \
      | sed -e 's/^#$//; s/^# //;'
}