Allow execute_query with app_name set to ml.app-name

[grtjn]

I changed execute_query_8 as shown below, and that allows me to at least invoke modules of current project with for instance:

  def replace_addresses
    country = find_arg(["--country"]) || "NL"
    lang = find_arg(["--language"]) || "Dutch"
    logger.info "Replacing customer addresses with #{country}/#{lang}.."
    r = execute_query(
      %Q{
        xquery version "1.0-ml";
        xdmp:invoke('/task/replace-addresses.xqy', (xs:QName("country"), "#{country}", xs:QName("lang"), "#{lang}"))
      },
      { :app_name => @properties["ml.app-name"] }
    )
    r.body = parse_body(r.body)
    logger.info r.body
  end

Patch for server_config.rb:

  def execute_query_8(query, properties = {})
    if properties[:app_name] != nil && properties[:app_name] != @properties["ml.app-name"]
      raise ExitException.new("Executing queries with an app_name (currently) not supported with ML8+")
    end

    headers = {
      "Content-Type" => "application/x-www-form-urlencoded"
    }

    params = {
      :xquery => query,
      :locale => LOCALE,
      :tzoffset => "-18000"
    }

    port = @qconsole_port
    if properties[:app_name] != nil
      port = @properties["ml.app-port"]
    end
    if properties[:db_name] != nil
      params[:database] = properties[:db_name]
    end

    r = go "#{@protocol}://#{@hostname}:#{port}/v1/eval", "post", headers, params

    raise ExitException.new(JSON.pretty_generate(JSON.parse(r.body))) if r.body.match(/\{"error"/)

    r
  end

[grtjn]

I marked it as bug as the missing app_name feature is pretty severe. We could further improve on this, by discovering the rest-port from app_name. I think I did something similar with capturing of app-builder projects, which are rest-based too..

[grtjn]

Even nicer would be having an invoke_query command, which takes modules-uri, and list of name/value pairs. Could make use of /v1/invoke, but would mean ML8+ support only. xcc invoke was never implemented from the looks of it..

[grtjn]

The above mentioned patch only works for rest and hybrid projects, and app-level auth might interfere as well. A more robust method is this:

  def execute_query_8(query, properties = {})
    if properties[:app_name] != nil && properties[:app_name] != @properties["ml.app-name"]
      raise ExitException.new("Executing queries with an app_name (currently) not supported with ML8+")
    end

    headers = {
      "Content-Type" => "application/x-www-form-urlencoded"
    }

    params = {
      :locale => LOCALE,
      :tzoffset => "-18000"
    }

    port = @qconsole_port
    if properties[:app_name] != nil
      params[:xquery] = %Q{
        xquery version "1.0-ml";
        let $query := <query><![CDATA[#{query}]]></query>
        return xdmp:eval(
          string($query),
          (),
          <options xmlns="xdmp:eval">
            <database>{xdmp:database("#{@properties["ml.content-db"]}")}</database>
            <modules>{xdmp:database("#{@properties["ml.modules-db"]}")}</modules>
          </options>
        )
      }
    else
      params[:xquery] = query
    end
    if properties[:db_name] != nil
      params[:database] = properties[:db_name]
    end

    r = go "#{@protocol}://#{@hostname}:#{port}/v1/eval", "post", headers, params

    raise ExitException.new(JSON.pretty_generate(JSON.parse(r.body))) if r.body.match(/\{"error"/)

    r
  end

[grtjn]

Tested above with this code:

  def test
    logger.info "Execute query with no params:"
    r = execute_query(
      %Q{
        xquery version "1.0-ml";

        "content-db: " || xdmp:database-name(xdmp:database()),
        "modules-db: " || xdmp:database-name(xdmp:modules-database()),
        "schemas-db: " || xdmp:database-name(xdmp:schema-database()),
        "trigger-db: " || xdmp:database-name(xdmp:triggers-database())
      }
    )
    r.body = parse_body r.body
    logger.info r.body
    logger.info ""

    logger.info "Execute query with db_name param:"
    r = execute_query(
      %Q{
        xquery version "1.0-ml";

        "content-db: " || xdmp:database-name(xdmp:database()),
        "modules-db: " || xdmp:database-name(xdmp:modules-database()),
        "schemas-db: " || xdmp:database-name(xdmp:schema-database()),
        "trigger-db: " || xdmp:database-name(xdmp:triggers-database())
      },
      { :db_name => @properties['ml.content-db'] }
    )
    r.body = parse_body r.body
    logger.info r.body
    logger.info ""

    logger.info "Execute query with app_name param:"
    r = execute_query(
      %Q{
        xquery version "1.0-ml";

        "content-db: " || xdmp:database-name(xdmp:database()),
        "modules-db: " || xdmp:database-name(xdmp:modules-database()),
        "schemas-db: " || xdmp:database-name(xdmp:schema-database()),
        "trigger-db: " || xdmp:database-name(xdmp:triggers-database())
      },
      { :app_name => @properties['ml.app-name'] }
    )
    r.body = parse_body r.body
    logger.info r.body
    logger.info ""
  end

[grtjn]

It should probably also take modules-root into account, but I leave that as exercise..

[grtjn]

We might also want to print a note that it might behave differently than it used to do if using app_name. A call to xdmp:server will return the id of App-Services, not of the app you specified. Not sure how the old xcc eval handled this, but I think it did better..

[grtjn]

Also, the eval negates the database param..

[grtjn]

Looking back through history, Roxy never used XCC eval, but has always relied on QC for eval in the past. My comment on xdmp:server returning wrong id has always applied, and can be neglected in that sense. It is a valid short-coming of Roxy, but one that has always existed.

And the eval indeed ignores the database param, but that is appropriate. One is supposed to call execute_query with either an app_name or a db_name, not both. If db_name is provided, the xdmp:eval is skipped, and query gets evaluated directly, while taking database param into account..

[grtjn]

Fixed in dev