1 % Embedded metasearching with the MasterKey Widget Set
9 There are lots of practical problems in building resource discovery
10 solutions. One of the biggest, and most ubiquitous is incorporating
11 metasearching functionality into existing web-sites -- for example,
12 content-management systems, library catalogues or intranets. In
13 general, even when access to core metasearching functionality is
14 provided by simple web-services such as
15 [Pazpar2](http://www.indexdata.com/pazpar2), integration work is seen
16 as a major part of most projects.
18 Index Data provides several different toolkits for communicating with
19 its metasearching middleware, trading off varying degrees of
20 flexibility against convenience:
22 * pz2.js -- a low-level JavaScript library for interrogating the
23 Service Proxy and Pazpar2. It allows the HTML/JavaScript programmer
24 to create JavaScript applications display facets, records, etc. that
25 are fetched from the metasearching middleware.
27 * masterkey-ui-core -- a higher-level, complex JavaScript library that
28 uses pz2.js to provide the pieces needed for building a
29 full-featured JavaScript application.
31 * MasterKey Demo UI -- an example of a searching application built on
32 top of masterkey-ui-core. Available as a public demo at
33 http://mk2.indexdata.com/
35 * MKDru -- a toolkit for embedding MasterKey-like searching into
38 All of these approaches require programming to a greater or lesser
39 extent. Against this backdrop, we introduced MKWS (the MasterKey
40 Widget Set) -- a set of simple, very high-level HTML+CSS+JavaScript
41 components that can be incorporated into any web-site to provide
42 MasterKey searching facilities. By placing `<div>`s with well-known
43 identifiers in any HTML page, the various components of an application
44 can be embedded: search-boxes, results areas, target information, etc.
50 The following is a complete MKWS-based searching application:
54 <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
55 <title>MKWS demo client</title>
56 <script type="text/javascript" src="http://mkws.indexdata.com/mkws-complete.js"></script>
57 <link rel="stylesheet" href="http://mkws.indexdata.com/mkws.css" />
60 <div id="mkwsSearch"></div>
61 <div id="mkwsResults"></div>
65 Go ahead, try it! You don't even need a web-server. Just copy and
66 paste this HTML into a file on your computer -- `/tmp/magic.html`,
67 say -- and point your web-browser at it:
68 `file:///tmp/magic.html`. Just like that, you have working
75 If you know any HTML, the structure of the file will be familar to
76 you: the `<html>` element at the top level contains a `<head>` and a
77 `<body>`. In addition to whatever else you might want to put on your
78 page, you can add MKWS elements.
80 These fall into two categories. First, the prerequisites in the HTML
81 header, which are loaded from the tool site mkws.indexdata.com:
84 contains all the JavaScript needed by the widget-set.
87 provides the default CSS styling
89 Second, within the HTML body, `<div>` elements with special IDs that
90 begin `mkws` can be provided. These are filled in by the MKWS code,
91 and provide the components of the searching UI. The very simple
92 application above has only two such components: a search box and a
93 results area. But more are supported. The main `<div>`s are:
95 * `mkwsSearch` -- provides the search box and button.
97 * `mkwsResults` -- provides the results area, including a list of
98 brief records (which open out into full versions when clicked),
99 paging for large results sets, facets for refining a search,
100 sorting facilities, etc.
102 * `mkwsLang` -- provides links to switch between one of several
103 different UI languages. By default, English, Danish and German are
106 * `mkwsSwitch` -- provides links to switch between a view of the
107 result records and of the targets that provide them. Only
108 meaningful when `mkwsTargets` is also provided.
110 * `mkwsTargets` -- the area where per-target information will appear
111 when selected by the link in the `mkwsSwitch` area. Of interest
112 mostly for fault diagnosis rather than for end-users.
114 * `mkwsStat` --provides a status line summarising the statistics of
117 To see all of these working together, just put them all into the HTML
120 <div id="mkwsSwitch"></div>
121 <div id="mkwsLang"></div>
122 <div id="mkwsSearch"></div>
123 <div id="mkwsResults"></div>
124 <div id="mkwsTargets"></div>
125 <div id="mkwsStat"></div>
131 Many aspects of the behaviour of MKWS can be modified by setting
132 parameters into the `mkws_config` object. **This must be done *before*
133 including the MKWS JavaScript** so that when that code is executed it
134 can refer to the configuration values. So the HTML header looks like
137 <script type="text/javascript">
140 sort_default: "title",
144 <script type="text/javascript" src="http://mkws.indexdata.com/mkws-complete.js"></script>
146 This configuration sets the UI language to Danish (rather than the
147 default of English), initially sorts search results by title rather
148 than relevance (though as always this can be changed in the UI) and
149 makes the search box a bit wider than the default.
151 The full set of supported configuration items is described in the
152 reference guide below.
155 Control over HTML and CSS
156 =========================
158 More sophisticated applications will not simply place the `<div>`s
159 together, but position them carefully within an existing page
160 framework -- such as a Drupal template, an OPAC or a SharePoint page.
162 While it's convenient for simple applications to use a monolithic
163 `mkwsResults` area which contains record, facets, sorting options,
164 etc., customised layouts may wish to treat each of these components
165 separately. In this case, `mkwsResults` can be omitted, and the
166 following lower-level components provided instead:
168 * `mkwsTermlists` -- provides the facets
170 * `mkwsRanking` -- provides the options for how records are sorted and
171 how many are included on each page of results.
173 * `mkwsPager` -- provides the links for navigating back and forth
174 through the pages of records.
176 * `mkwsNavi` -- when a search result has been narrowed by one or more
177 facets, this area shows the names of those facets, and allows the
178 selected values to be clicked in order to remove them.
180 * `mkwsRecords` -- lists the actual result records.
182 Customisation of MKWS searching widgets can also be achieved by
183 overriding the styles set in the toolkit's CSS stylesheet. The default
184 styles can be inspected in `mkws.css` and overridden in any
185 styles that appears later in the HTML than that file. At the simplest
186 level, this might just mean changing fonts, sizes and colours, but
187 more fundamental changes are also possible.
189 To properly apply styles, it's necessary to understand how the HTML is
190 structured, e.g. which elements are nested within which
191 containers. The structures used by the widget-set are described in the
192 reference guide below.
202 Some applications might like to open with content in the area that
203 will subsequently be filled with result-records -- a message of the
204 day, a welcome message or a help page. This can be done by placing an
205 `mkwsMOTD` division anywhere on the page. It will be moved into the
206 `mkwsResults` area and initially displayed, but will be hidden when a
210 Customised display using Handlebars templates
211 ---------------------------------------------
213 Certain aspects of the widget-set's display can be customised by
214 providing Handlebars templates with well-known classes that begin with
215 the string `mkwsTemplate_`. At present, the supported templates are:
217 * `mkwsTemplate_Summary` -- used for each summary record in a list of
220 * `mkwsTemplate_Record` -- used when displaying a full record.
222 For both of these the metadata record is passed in, and its fields can
223 be referenced in the template. As well as the metadata fields
224 (`md-*`), two special fields are provided to the `mkwsTemplate_Summary`
225 template, for creating popup links for full records. These are `_id`,
226 which must be provided as the `id` attribute of a link tag, and
227 `_onclick`, which must be provided as the `onclick` attribute.
229 For example, an application can install a simple author+title summary
230 record in place of the usual one providing the following template:
232 <script class="mkwsTemplate_Summary" type="text/x-handlebars-template">
234 <span>{{md-author}}</span>
236 <a href="#" id="{{_id}}" onclick="{{_onclick}}">
241 For details of Handlebars template syntax, see
242 [the online documentation](http://handlebarsjs.com/).
248 Metasearching applications may need to appear differently on
249 small-screened mobile devices, or change their appearance when
250 screen-width changes (as when a small device is rotated). To achieve
251 this, MKWS supports responsive design which will move the termlists to
252 the bottom on narrow screens and to the sidebar on wide screens.
254 To turn on this behaviour, set the `responsive_design_width` to the desired
255 threshhold width in pixels. For example:
257 <script type="text/javascript">
259 responsive_design_width: 990
263 If individual result-related components are in use in place of the
264 all-in-one mkwsResults, then the redesigned application needs to
265 specify the locations where the termlists should appear in both
266 cases. In this case, wrap the wide-screen `mkwsTermlists` element in a
267 `mkwsTermlists-Container-wide` element; and provide an
268 `mkwsTermlists-Container-narrow` element in the place where the narrow-screen
269 termlists should appear.
272 Popup results with jQuery UI
273 ----------------------------
275 The [jQuery UI library](http://en.wikipedia.org/wiki/JQuery_UI)
276 can be used to construct MKWS applications in which the only component
277 generally visible on the page is a search box, and the results appear
278 in a popup. The key part of such an application is this invocation of
279 the MKWS jQuery plugin:
281 <div class="mkwsSearch"></div>
282 <div class="mkwsPopup" popup_width="1024" popup_height="650" popup_modal="0" popup_autoOpen="0" popup_button="input.mkwsButton">
283 <div class="mkwsSwitch"></div>
284 <div class="mkwsLang"></div>
285 <div class="mkwsResults"></div>
286 <div class="mkwsTargets"></div>
287 <div class="mkwsStat"></div>
290 The necessary scaffolding can be seen in an example application,
291 http://example.indexdata.com/index-popup.html
294 Authentication and target configuration
295 ---------------------------------------
297 By default, MKWS configures itself to use a demonstration account on a
298 service hosted by mkws.indexdata.com. This account (username `demo`,
299 password `demo`) provides access to about a dozen free data
300 sources. Authentication onto this service is via an authentication URL
301 on the same MKWS server, so no explicit configuration is needed.
303 In order to search in a customised set of targets, including
304 subscription resources, it's necessary to create an account with
305 Index Data's hosted service proxy, and protect that account with
306 authentication tokens (to prevent unauthorised use of subscription
307 resources). For information on how to do this, see the next section.
310 MKWS Target Selection
311 =====================
313 MKWS accesses targets using the Pazpar2 metasearching engine. Although
314 Pazpar2 can be used directly, using a statically configured set of
315 targets, this usage is unusual. More often, Pazpar2 is fronted by the
316 Service Proxy (SP), which manages authentication, sessions, target
319 This document assumes the SP is used, and explains how to go about
320 making a set of targets (a "library") available, how to connect your
321 MKWS application to that library, and how to choose which of the
322 available targets to use.
325 Maintaining the library
326 -----------------------
328 The service proxy accesses sets of targets that are known as
329 "libraries". In general, each customer will have their own library,
330 though some standard libraries may be shared between many customers --
331 for example, a library containing all open-access academic journals.
332 A library can also contain other configuration information, including
333 the set of categories by which targets are classified for the library.
335 Libraries are maintained using MKAdmin (MasterKey
336 Admin). Specifically, those used by MKWS are generally maintained on
337 the "MKX Admin" installation at
338 <http://mkx-admin.indexdata.com/console/>
340 In general, Index Data will create a library for each customer, then
341 give the customer a username/password pair that they can use to enter
342 MKAdmin and administrate that library.
344 Once logged in, customers can select which targets to include (from
345 the list of several thousand that MKAdmin knows about), and make
346 customer-specific modifications -- e.g. overriding the titles of the
349 Most importantly, customers' administrators can add authentication
350 credentials that the Service Proxy will used on their behalf when
351 accessing subscription resources -- username/password pairs or proxies
352 to use for IP-based authentication. Note that **it is then crucial to
353 secure the library from use by unauthorised clients**, otherwise the
354 customer's paid subscriptions will be exploited.
356 Access to libraries is managed by creating one or more "User Access"
357 records in MKAdmin, under the tab of that name. Each of these records
358 provides a combination of credentials and other data that allow an
359 incoming MKWS client to be identified as having legitimate access to
360 the library. The authentication process, described below, works by
361 searching for a matching User Access record.
364 Authenticating your MWKS application onto the library
365 -----------------------------------------------------
367 Some MKWS applications will be content to use the default library with
368 its selection of targets. Most, though, will want to define their own
369 library providing a different range of available targets. An important
370 case is that of applications that authenticate onto subscription
371 resources by means of back-end site credentials stored in MKAdmin:
372 precautions must be taken so that such library accounts do not allow
375 Setting up such a library is a process of several stages.
377 ### Create the User Access account
379 Log in to MKAdmin to add a User Access account for your library:
381 * Go to <http://mkx-admin.indexdata.com/console/>
382 * Enter the adminstrative username/password
383 * Go to the User Access tab
384 * Create an end-user account
385 * Depending on what authentication method it be used, set the
386 User Access account's username and password, or referring URL, or
387 Service Proxy hostname, or IP-address range.
389 If your MWKS application runs at a well-known, permanent address --
390 <http://yourname.com/app.html>, say -- you can set the User Access
391 record so that this originating URL is recognised by setting it into
392 the "Referring URL" field.
394 If your application accesses the Service Proxy by a unique virtual
395 hostname -- yourname.sp-mkws.indexdata.com, say -- you can tie the use
396 of this hostname to your library by setting the User Access record's
397 "Host Name" field to name of the host where the SP is accessed. **Note
398 that this is not secure, as other applications can use this virtual
399 hostname to gain access to your library.**
401 > TODO Authentication by IP address does not yet work correctly -- see
402 > bug MKWS-234 ("Improve SP configuration/proxying for better
405 Alternatively, your application can authenticate by username and
406 password credentials. This is a useful approach in several situations,
407 including when you need to specify the use of a different library from
408 usual one. To arrange for this, set the username and password as a
409 single string separated by a slash -- e.g. "mike/swordfish" -- into
410 the User Access record's Authentication field.
412 You can create multiple User Access records: for example, one that
413 uses Referring URL, and another that uses a username/password pair to
414 be used when running an application from a different URL.
416 ### Tell the application to use the library
418 In the HTML of the application, tell MKWS to authenticate on to the
419 Service Proxy. When IP-based, referer-based or hostname-based
420 authentication is used, this is very simple:
422 <script type="text/javascript">
423 var mkws_config = { service_proxy_auth:
424 "//sp-mkws.indexdata.com/service-proxy/?command=auth&action=perconfig" };
427 > TODO This should be the default setting
429 And ensure that access to the MWKS application is from the correct
430 Referrer URL or IP-range.
432 ### (Optional): access by a different virtual hostname
434 When hostname-based authentication is in use, it's necessary to access
435 the Service Proxy as the correctly named virtual host. This can be
436 done by setting the `service_proxy_auth` configuration item to a
437 URL containing that hostname, such as
438 <//yourname.sp-mkws.indexdata.com/service-proxy/?command=auth&action=perconfig>
440 > TODO It should be possible to change just the hostname without
441 > needing to repeat the rest of the URL (protocol, path, query)
443 > TODO When changing the SP authentication URL, the Pazpar2 URL should
444 > in general change along with it.
446 ### (Optional): embed credentials for access to the library
448 When credential-based authentication is in use (username and
449 password), it's necessary to pass these credentials into the Service
450 Proxy when establishing the session. This can most simply be done just
451 by setting the `service_proxy_auth` configuration item to a URL such as
452 <//sp-mkws.indexdata.com/service-proxy/?command=auth&action=perconfig&username=mike&password=swordfish>
454 > TODO It should be possible to add the username and password to the
455 > configuration without needing to repeat the rest of the URL.
457 ### (Optional): conceal credentials from HTML source
459 Using a credential-based Service-Proxy authentication URL such as the
460 one above reveals the the credentials to public view -- to anyone who
461 does View Source on the MKWS application. This may be acceptable for
462 some libraries, but is intolerable for those which provide
463 authenticated access to subscription resources.
465 In these circumstances, a more elaborate approach is necessary. The
466 idea is to make a URL local to the customer that is used for
467 authentication onto the Service Proxy, hiding the credentials in a
468 local rewrite rule. Then local mechanisms can be used to limit access
469 to that local authentication URL. Here is one way to do it when
470 Apache2 is the application's web-server, which we will call
473 - Add a rewriting authentication alias to the configuration:
476 RewriteRule /spauth/ http://mkws.indexdata.com/service-proxy/?command=auth&action=check,login&username=U&password=PW [P]
478 - Set the MKWS configuration item `service_proxy_auth` to
479 <http://yourname.com/spauth/>
480 - Protect access to the local path <http://yourname.com/spauth/>
481 (e.g. using a .htaccess file).
484 Choosing targets from the library
485 ---------------------------------
487 MKWS applications can choose what subset of the library's targets to
488 use, by means of several alternative settings on individual widgets or
489 in the `mkws_config` structure:
491 * `targets` -- contains a Pazpar2 targets string, typically of the form
492 "pz:id=" or "pz:id~" followed by a pipe-separated list of low-level
494 At present, these IDs can take one of two forms, depending on the
495 configuration of the Service Proxy being used: they may be based on
496 ZURLs (so a typical value would be something like
497 `pz:id=josiah.brown.edu:210/innopac|lui.indexdata.com:8080/solr4/select?fq=database:4902`)
498 or they may be UDBs (so a typical value would be something like
499 `pz:id=brown|artstor`)
501 * `targetfilter` -- contains a CQL query which is used to find relevant
502 targets from the relvant library. For example,
507 * `target` -- contains a single UDB, that of the sole target to be
510 This is merely syntactic sugar for "targetfilter" with the query
513 For example, a `Records` widget can be limited to searching only in
514 targets that have been categorised as news sources by providing an
515 attribute as follows:
517 <div class="mkwsRecords" targetfilter='categories=news'/>
526 The configuration object `mkws_config` may be created before including
527 the MKWS JavaScript code to modify default behaviour. This structure
528 is a key-value lookup table, whose entries are described in the table
529 below. All entries are optional, but if specified must be given values
530 of the specified type. If ommitted, each setting takes the indicated
531 default value; long default values are in footnotes to keep the table
535 Element Type Default Description
536 -------- ----- --------- ------------
537 debug_level int 1 Level of debugging output to emit. 0 = none, 1 = messages, 2 = messages with
538 datestamps, 3 = messages with datestamps and stack-traces.
540 facets array *Note 1* Ordered list of names of facets to display. Supported facet names are
541 `xtargets`, `subject` and `author`.
543 lang string en Code of the default language to display the UI in. Supported language codes are `en` =
544 English, `de` = German, `da` = Danish, and whatever additional languages are configured
545 using `language_*` entries (see below).
547 lang_options array [] A list of the languages to offer as options. If empty (the default), then all
548 configured languages are listed.
550 language_* hash Support for any number of languages can be added by providing entries whose name is
551 `language_` followed by the code of the language. See the separate section below for
554 pazpar2_url string *Note 2* The URL used to access the metasearch middleware. This service must be configured to
555 provide search results, facets, etc. It may be either unmediated or Pazpar2 the
556 MasterKey Service Proxy, which mediates access to an underlying Pazpar2 instance. In
557 the latter case, `service_proxy_auth` must be provided.
559 perpage_default string 20 The initial value for the number of records to show on each page.
561 perpage_options array *Note 3* A list of candidate page sizes. Users can choose between these to determine how many
562 records are displayed on each page of results.
564 query_width int 50 The width of the query box, in characters.
566 responsive_design_width int If defined, then the facets display moves between two locations as the screen-width
567 varies, as described above. The specified number is the threshhold width, in pixels,
568 at which the facets move between their two locations.
570 service_proxy_auth url *Note 4* A URL which, when `use_service_proxy` is true, is fetched once at the beginning of each
571 session to authenticate the user and establish a session that encompasses a defined set
572 of targets to search in.
574 service_proxy_auth_domain domain Can be set to the domain for which `service_proxy_auth` proxies authentication, so
575 that cookies are rewritten to appear to be from this domain. In general, this is not
576 necessary, as this setting defaults to the domain of `pazpar2_url`.
578 show_lang bool true Indicates whether or not to display the language menu.
580 show_perpage bool true Indicates whether or not to display the perpage menu.
582 show_sort bool true Indicates whether or not to display the sort menu.
584 show_switch bool true Indicates whether or not to display the switch menu, for switching between showing
585 retrieved records and target information.
587 sort_default string relevance The label of the default sort criterion to use. Must be one of those in the `sort`
590 sort_options array *Note 6* List of supported sort criteria. Each element of the list is itself a two-element list:
591 the first element of each sublist is a pazpar2 sort-expression such as `data:0` and
592 the second is a human-readable label such as `newest`.
594 use_service_proxy bool true If true, then a Service Proxy is used to deliver searching services rather than raw
598 Perhaps we should get rid of the `show_lang`, `show_perpage`,
599 `show_sort` and `show_switch` configuration items, and simply display the relevant menus
600 only when their containers are provided -- e.g. an `mkwsLang` element
601 for the language menu. But for now we retain these, as an easier route
602 to lightly customise the display than my changing providing a full HTML
607 1. ["sources", "subjects", "authors"]
609 2. /pazpar2/search.pz2
613 4. http://mkws.indexdata.com/service-proxy-auth
615 5. http://mkws.indexdata.com/service-proxy/
617 6. [["relevance"], ["title:1", "title"], ["date:0", "newest"], ["date:1", "oldest"]]
620 Language specification
621 ----------------------
623 Support for another UI language can be added by providing an entry in
624 the `mkws_config` object whose name is `language_` followed by the
625 name of the language: for example, `language_French` to support
626 French. Then value of this entry must be a key-value lookup table,
627 mapping the English-language strings of the UI into their equivalents
628 in the specified language. For example:
632 "Authors": "Auteurs",
633 "Subjects": "Sujets",
634 // ... and others ...
638 The following strings occurring in the UI can be translated:
654 In addition, facet names can be translated:
660 Finally, the names of fields in the full-record display can be
661 translated. These include, but may not be limited to:
671 jQuery UI popup invocation
672 --------------------------
674 The MasterKey Widget Set can be invoked in a popup window on top of the page.
676 Note that when using the `popup` layout, facilities from the jQuery UI
677 toolkit are used, so it's necessary to include both CSS and JavaScript
678 from that toolkit. The relevant lines are:
680 <script src="http://code.jquery.com/ui/1.10.3/jquery-ui.min.js"></script>
681 <link rel="stylesheet" type="text/css"
682 href="http://code.jquery.com/ui/1.10.3/themes/smoothness/jquery-ui.css" />
684 <div class="mkwsSearch"></div>
685 <div class="mkwsPopup" popup_width="1024" popup_height="650" popup_modal="0" popup_autoOpen="0" popup_button="input.mkwsButton">
686 <div class="mkwsSwitch"></div>
687 <div class="mkwsLang"></div>
688 <div class="mkwsResults"></div>
689 <div class="mkwsTargets"></div>
690 <div class="mkwsStat"></div>
694 Element Type Default Description
695 -------- ----- --------- ------------
696 popup_width string 880 Width of the popup window (if used), in
699 popup_height string 760 Height of the popup window (if used), in
702 popup_button string input.mkwsButton (Never change this.)
704 popup_modal string 0 Modal confirmation mode. Valid values are 0 or 1
706 popup_autoOpen string 1 Open popup window on load. Valid values are 0 or 1
711 The structure of the HTML generated by the MKWS widgets
712 -------------------------------------------------------
714 In order to override the default CSS styles provided by the MasterKey Widget
715 Set, it's necessary to understand that structure of the HTML elements that are
716 generated within the components. This knowledge make it possible, for example,
717 to style each `<div>` with class `term` but only when it occurs inside an
718 element with ID `#mkwsTermlists`, so as to avoid inadvertently styling other
719 elements using the same class in the non-MKWS parts of the page.
721 The HTML structure is as follows. As in CSS, #ID indicates a unique identifier
722 and .CLASS indicates an instance of a class.
732 input#mkwsQuery type=text
733 input#mkwsButton type=submit
736 (no contents -- used only for masking)
757 span (for sequence number)
759 span (for other information such as author)
760 div.details (sometimes)
783 Copyright (C) 2013-2014 by IndexData ApS, <http://www.indexdata.com>