Using the WorldCat Search API

# let's load this package before getting started
#> Loading required package: data.table

Intermediate knowledge of R is required to follow the examples here.

Basic knowledge of how to manipulate data.tables if helpful but following the examples should be possible without it.


libbib’s ability to communicate to the WorldCat Search API is probably the most helpful capability of the package. As such, an entire vignette dedicated to this powerful tool is warranted.

Also of note is that documentation on how to use this API is scattered throughout the web, in different places, and sometimes even no longer available and only accessible through the internet archive. Because of this, the goal of this vignette is not only to document libbib’s worldcat_api_search function, but to provide a single location for compiling this archived documentation and providing information/examples of what the WorldCat Search API is capable, in general.

The terminology/nomenclature of the different WorldCat API offerings is fuzzy and inconsistent. The specific kind of Search API that this package offers usage of can, more specifically, be described as the ‘WorldCat SRU Search API version 1’.

There is a version 2 of this API, but version 1 hasn’t been sunsetted as of yet.

First, let’s speak of what this API is and what it is not.

What the WorldCat SRU Search API is

This API is allows a developer to search for bibliographic records that are cataloged in WorldCat.

The queries are made using the SRU standard search protocol (Search/Retrieve via URL) using a standard query syntax called CQL (Contextual Query Language). The WorldCat Search API doesn’t implement all features of CQL; this vignette will illustrate SRU/CQL only insofar as it is supported by this API.

This API is open to libraries that maintain both WorldCat Discovery and OCLC Cataloging subscriptions and needs an API key (called a WSKey) to work. A request can be made for a key (if your institution doesn’t already have one) via this link.

What the WorldCat SRU Search API is not

This API is not the OpenSearch/Basic API, which doesn’t allow field-specific searches and only supports keyword-anywhere searches.

This is also not the same as using the advanced search option on That option only provides a subset of the bibliographic records in WorldCat and is a far less powerful tool.

This API is most akin to using the “expert search” option in OCLC’s FirstSearch, using the WorldCat database, but differs in that (a) searching the API is programmatically automate-able, and (b) the API allows for, still, more powerful search queries.

What libbib’s worldcat_api_search function provides

At its most basic, this function takes a SRU query and returns a data.table with most of the bibliographic metadata from the MARCXml that the API returns. We’ll see how the behavior of this function can be controlled by specifying certain function parameters.

This function also offers assistance with the SRU query syntax. Mainly, the function allows you to substitute the arcane search index codes for more human-readable equivalents, prefixed be a (US) dollar-sign. Examples of these aids will be explained later in the vignette.

An example of usage and what the function returns

Let’s look at an example of a simple query and what the returned data.table looks like.

Specifically, we’ll search for “Madame Bovary” by “Gustave Flaubert” and we’ll only show the first three results

library(libbib)      # load this package

result <- worldcat_api_search('$title="Madame Bovary" and
                                 $author="Gustave Flaubert"')

# get the column names
#>  [1] "total_wc_results" "result_number"   "oclc"           "isbn"            
#>  [5] "issn"             "title"           "author"         "pub_date"        
#>  [9] "lang_code"        "bib_level"       "record_type"    "pub_place_code"  
#> [13] "publisher"        "leader"          "oh08"           "query"

# show the first three results
#>    total_wc_results result_number       oclc          isbn   issn
#>              <char>         <int>     <char>        <char> <char>
#> 1:              986             1 1125170419 9782253183464   <NA>
#> 2:              986             2 1049849403 9788415618843   <NA>
#> 3:              986             3 1203070641 9781664921993   <NA>
#>              title             author pub_date lang_code      bib_level
#>             <char>             <char>    <int>    <char>         <char>
#> 1: Madame Bovary : Flaubert, Gustave,     2019       fre Monograph/Item
#> 2: Madame Bovary / Flaubert, Gustave,     2019       spa Monograph/Item
#> 3: Madame Bovary / Flaubert, Gustave,     2021       eng Monograph/Item
#>                   record_type pub_place_code          publisher
#>                        <char>         <char>             <char>
#> 1:          Language Material             fr le Livre de poche,
#> 2:          Language Material             sp               <NA>
#> 3: Nonmusical sound recording            ohu               <NA>
#>                      leader                                     oh08
#>                      <char>                                   <char>
#> 1: 00000cam a2200000Mi 4500 190619s2019    fr a   g      000 1 fre d
#> 2: 00000cam a2200000Ii 4500 180827t20192018sp a          000 1 spa d
#> 3: 00000cim a2200000Mi 4500 201104s2021    ohunnnneq      f  n eng d
#>                                                    query
#>                                                   <char>
#> 1: srw.ti="Madame Bovary" and"Gustave Flauber...
#> 2: srw.ti="Madame Bovary" and"Gustave Flauber...
#> 3: srw.ti="Madame Bovary" and"Gustave Flauber...

This should give you an idea of the rich information returned by the results data.table. All of the information returned is from (or derived from) the MARCXml that the API returns, save for

Hereafter, the output of the example queries will be abbreviated, truncated, elided, or only show a subset of columns in order to save space and aid in following the guide.

What does an error look like

If you craft a query that yields no results, a message telling you such will be displayed, and the return value is NULL.

If you make an error in the query syntax, no results will be returned, but a diagnostic message returned from the server may tell you what went wrong. Here are two examples…

# missing ending double quotes
worldcat_api_search('$title="Madame Bovary and $langauge=greek')
#> Received diagnostic message: Query syntax error (org.z3950.zing.cql.CQLParseException: expected index or term, got EOF)
#> no results found

# "$titley" is not a valid search index
worldcat_api_search('$titley="Madame Bovary" and $langauge=greek')
#> Received diagnostic message: Unsupported index (srw.tiy)
#> no results found

#> Received diagnostic message: Limit index, can only be used to narrow a result
#> for a non-limit index (
#> no results found

This last one failed because $holding_library is considered a limit index. This means that you can’t use it to search directly, you can only combine with search facet with other non-limiting indexes to filter the results. Hopefully this will make more sense as you go through this vignette.

worldcat_api_search by example / quick start

In guides like this, there is frequently tension between wanting to be a complete reference, and wanting to show cool/helpful examples at a glance.

In service of easing this tension, we’ll first look at some examples illustrating things this function can do and then turn our attention to an adequate explanation of boolean operators, relations operators, search indexes, etc…

These examples are largely taken from the examples section of the R documentation for this function.

These examples (and the ones further along in this vignette) will also use data.table syntax to limit number of rows and columns returned to aid reading.

We’ll be focusing on the types of queries you can use in this section; controlling the behavior of the function via changing parameters will come later.


# title search for "The Brothers Karamazov"
results <- worldcat_api_search('$title="Brothers Karamazov"')

# Madame Bovary by Gustave Flaubert in Greek
sru <- '$author="Gustave Flaubert" and
          $title="Madame Bovary" and
results <- worldcat_api_search(sru)

# Hip Hop materials on wax, cassette, or CD published from
# 1987 to 1990
sru <- '(($material_type=cas or $material_type=cda or $material_type=lps)
           and $subject="Rap") and $year="1987-1990"'
results <- worldcat_api_search(sru)

# keyword search for "Common Lisp" for materials held
# at The New York Public Library
sru <- '$keyword="common lisp" and $holding_library=NYP'
results <- worldcat_api_search(sru)

# keyword search for "Common Lisp" for materials held
# by any of the members of the "Manhattan Research
# Library Initiative" (MaRLI) joint borrowing program
# (New York Public Library, Columbia University, and
# New York University)
sru <- '($keyword="common lisp" and $holding_library=NYP)
          or ($keyword="common lisp" and $holding_library=ZYU)
          or ($keyword="common lisp" and $holding_library=ZCU)'
results <- worldcat_api_search(sru)

# Books (only books) about Ethics (by dewey division 170s or
# LC call number subject class "BJ") published in the 19th
# century
sru <- '($dewey="17*" or $lc_call="bj*") and $year="18*" and
results <- worldcat_api_search(sru)

# Materials on Musicology (by Dewey division 780s) at
# the New York Public Library and not held by any
# other insitution
sru <- '$dewey="78*" and $holding_library=NYP and
results <- worldcat_api_search(sru)

# Search for materials on "Danger Music" published since 2010
results <- worldcat_api_search('$keyword="danger music" and $year="2010-"')

Now that we’ve seen these examples, sans explanation, we can now have a closer look into the components of a query. Broadly speaking, there are four concepts to be aware of…

We’ll be looking at each in this order, because I think that makes the most sense. To illustrate the first three concepts, though, we have to use search indexes, before a formal explanation of what they are.

Briefly, a search index is a facet along which to search. In all of the examples above, the search indexes were prefixed by a $ character (e.g. $title, author, $keyword, etc…)

Relations operators

There are four relations operators available for use…

It should be noted that not every relations operator is available for use with every search index.


This was the most common operator used in the quick-start examples above. Though this operator can be most fully understood via contrast with the next operator, exact, suffice it to say, for now, that using = means that all of your search terms must match, without intervening words. This is sometimes referred to as an “un-anchored” search.

If the phrase you’re search for has spaces in it, you have to surround it with double quotes. Since the SRU query to the worldcat_api_search function must be a string, and strings can be made with single quotes and double quotes, we need to surround the entire query with single quotes.

We can get around this by “escaping” the quotes, but using only single quotes to surround the whole query is the most elegant approach.

You can use both spaces and + to separate the search index and the search term on both sides of the relations operator…

sru <- '$title+=+"Common Lisp"'
results <- worldcat_api_search(sru)
results[1:3, .(total_wc_results, title)]
#>    total_wc_results                   title
#>              <char>                  <char>
#> 1:              367           Common Lisp /
#> 2:              367 Practical Common Lisp /
#> 3:              367           Common LISP /

is the same as…

sru <- '$title = "Common Lisp"'
results <- worldcat_api_search(sru)
results[1:3, .(total_wc_results, title)]
#>    total_wc_results                   title
#>              <char>                  <char>
#> 1:              367           Common Lisp /
#> 2:              367 Practical Common Lisp /
#> 3:              367           Common LISP /

But we’ll be using spaces here.

The = relations operator is the only one that doesn’t require any space or + on either side of the operator, and we’ll be using both the spaced version and non-spaced version in these examples for this operator, only.


Using any means that any of your search terms (inside the double quotes) can match. For example…

results <- worldcat_api_search('$title any "Common Lisp"')
results[1:3, .(total_wc_results, title)]
#>    total_wc_results                                                 title
#>              <char>                                                <char>
#> 1:           351777 An Inquiry into the Human Mind on the Principles o...
#> 2:           351777                            The book of common prayer.
#> 3:           351777                        Dictionary of Phrase and Fable

matches the title “The book of common prayer” (decidedly not a book of the subject of Common Lisp) since it has the word “common” in it. (Note that capitalization doesn’t matter).

Using any in this context is tantamount to using an or boolean operator, which we’ll look at in the next section…

results <- worldcat_api_search('$title = Common or $title = "Lisp"')
results[1:3, .(total_wc_results, title)]
#>    total_wc_results                                                 title
#>              <char>                                                <char>
#> 1:           351777 An Inquiry into the Human Mind on the Principles o...
#> 2:           351777                            The book of common prayer.
#> 3:           351777                        Dictionary of Phrase and Fable

(Note the same number and order of results)


Using the all operator means that all of the search terms must match but the search terms can be in any order and have intervening terms in between. For example, search like $title all "Common Lisp" can match a (fictional) book named “Speaking with a lisp is common”.

A quick note before continuing

If your search term is one word, you do not have to use double quotes to surround the term. For example, $title = Ethics and $title exact Ethics work perfectly well.

If your search term as a single quote in it, it must be escaped so that R doesn’t interpret it as the end of the search string. That being said, it appears as if you can just drop the single quote and the search will carry on perfectly fine…

results <- worldcat_api_search('$title exact "Finnegans Wake"')
results[1:3, .(total_wc_results, title, query)]
#>    total_wc_results            title                         query
#>              <char>           <char>                        <char>
#> 1:              761 Finnegans wake / srw.ti exact "Finnegans Wake"
#> 2:              761   Finnegans Wake srw.ti exact "Finnegans Wake"
#> 3:              761   Finnegans wake srw.ti exact "Finnegans Wake"

# yields the same results as

results <- worldcat_api_search('$title exact "Finnegan\'s Wake"')
results[1:3, .(total_wc_results, title, query)]
#>    total_wc_results            title                         query
#>              <char>           <char>                        <char>
#> 1:              761 Finnegans wake / srw.ti exact "Finnegan's Wake"
#> 2:              761   Finnegans Wake srw.ti exact "Finnegan's Wake"
#> 3:              761   Finnegans wake srw.ti exact "Finnegan's Wake"

I suspect single quotes are automatically elided by the API.

Boolean operators (and operator precedence)

As evinced in the examples shown earlier in this vignette, you can use the boolean operators and, or, and not to refine your search.

As you mix different boolean operators in a single query, care must be taken to ensure that the order/precedence of these operators matches your intention.

For example, in one of the examples above, we searched for Hip Hop materials on wax, cassette, or CD published from 1987 to 1990 with the following SRU query

'(($material_type=cas or $material_type=cda or $material_type=lps)
   and $subject="Rap") and $year="1987-1990"'

It’s important to note that if we wrote the query like shown below

$material_type=cas or $material_type=cda or $material_type=lps
   and $subject="Rap" and $year="1987-1990"'

without any parentheses, the meaning of the query would be ambiguous. Make sure you use parentheses around distinct sections of your search incantation to disambiguate the query.


taken verbatim from the archived documentation

For right truncation use an asterisk - * There must be at least three characters before the * for the query to work. There is no left truncation.

To wildcard a single character wildcard use a number sign - #. So a query for wom#n provides results that include both woman and women in the results.

For a 0-9 number of characters as wildcard characters use ?n. So a query for colo?1r provides results of color and colour.

To wildcard characters within (or at the end) a term use the question mark - ?. So a query for colo?r provides results of color, colour, colonizer, and colorimeter.

Table of all available search indexes

Formal search index code libbib alias $keyword
srw.ti $title
srw.ln $language $author
srw.yr $year $subject $holding_library $material_type $oclc $lc_call
srw.dd $dewey
srw.dn $lccn $isbn $issn $library_holdings_group $language_code $place_of_publication
srw.pb $publisher $access_method $corporate_conference_name
srw.pc $dlc_limit
srw.dt $document_type $government_document_number $music_publisher_number
srw.nt $notes
srw.on $open_digital_limit $personal_name $series $standard_number

Selected Search Indexes in focus

$title / srw.ti

According to the archived Search API documentation, (I have strong doubts) the title search phrase will automatically elide certain common and “un-important” words. In the field of Natural Language Processing, we call these “stop words”.

The archive docs indicates that the following words will be removed from the search phase:

a, als, am, an, are, as, at, auf, aus, be, but, by, das, dass, de, der, des, dich, dir, du, er, es, for, from, had, have, he, her, his, how, ihr, ihre, ihres, im, in, is, ist, it, kein, la, le, les, mein, mich, mir, mit, of, on, sein, sie, that, the, this, to, un, une, von, was, wer, which, wie, wird, with, yousie, that, the, this, to, un, une, von, was, wer, which, wie, wird, with, you.

This means that title searches for The Brothers Karamazov, La Noche Boca Arriba, Das Spiel ist aus, or Das Kapital will, alledgely, internally use the seach phrases Brothers Karamazov, Noche Boca Arriba, Spiel, and Kapital, respectively.

When searching the title index, though, I’d leave these words in the SRU search query for two reasons…

$holding_library /

As shown in one of the examples above, the holding library search term is the official OCLC designator. You can search for an institution’s code by the institutions name using this link.

$material_type /

An exhaustive crosswalk of all material types and their (normally 3-letter) codes would be too large to include here, but you can access the crosswalk in the documentation that is archived at this link

$library_holdings_group /

The “library holdings group” search term is, perhaps, a confusing one. The table below is a crosswalk between the search codes to use in the SRU query, and what they mean.

Search code Meaning
05 5 or more holdings
06 10 or more holdings
07 50 or more holdings
08 100 or more holdings
09 500 or more holdings
10 No holdings
11 1 holding only
12 2 – 4 holdings
13 5 – 9 holdings
14 10 – 24 holdings
15 25 – 49 holdings
16 50 - 74 holdings
17 75 – 99 holdings
18 100 - 149 holdings
19 150 - 199 holdings
20 200 - 299 holdings
21 300 - 399 holdings
22 400 - 499 holdings
23 500 - 599 holdings
24 600 - 699 holdings
25 700 - 799 holdings
26 800 - 899 holdings
27 900 - 999 holdings
28 1,000 - 1,499 holdings
29 1,500 - 1,999 holdings
30 2,000 - 2,499 holdings
31 2,500 or more holdings

So, for example, if you wanted to limit your search to items that are held by between 100 to 149 institutions, you would have to add $library_holdings_group=18 to your SRU query.


For all the available search indexes, it is very helpful to know which exact MARC fields are searched in each one. This archived documentation link contains that information.

On that page, the SRU index codes have a prefix of sru, but it’s really srw. Consult the Table of all available search indexes above to find the libbib aliases for the indexes of interest (though, of course, you can use the un-translated index codes if you’d like, too).

Options in the worldcat_api_search function

Besides for, of course, the SRU query, the worldcat_api_search function takes a number of optional parameters that can be used to alter its semantics. Below is a list and explanation of each of those parameters. (This information is also available by running help("worldcat_api_search") in an R console after loading the libbib package.)

Combining this with the other API search functions in libbib

There are numerous ways to combine the worldcat_api_search with the other functions that libbib provides to do some really useful investigations. In the example below, will be using the worldcat_api_search and worldcat_api_locations_by_oclc functions to get a list of institutions that hold any edition of my textbook. (This example uses some data.table specific syntax for brevity, but it will work with base R [or “tidyverse”] translations just fine.)

First, let’s use the search function to get a list of all search results for the book…

results <- worldcat_api_search('srw.ti="Data Analysis with r"

# inspect some of the columns in the first 5 results
results[1:5, .(total_wc_results, result_number, oclc,
               title, author, pub_date)
#>    total_wc_results result_number       oclc                   title
#>              <char>         <int>     <char>                  <char>
#> 1:               11             1 1005106045 DATA ANALYSIS WITH R -.
#> 2:               11             2 1089176194  Data analysis with R :
#> 3:               11             3  949229431  Data analysis with R :
#> 4:               11             4 1242682069    Data Analysis with R
#> 5:               11             5 1242707288    Data Analysis with R
#>              author pub_date
#>              <char>    <int>
#> 1: FISCHETTI, TONY.     2018
#> 2: Fischetti, Tony.     2018
#> 3: Fischetti, Tony.     2015
#> 4:  Fischetti, Tony     2015
#> 5:  Fischetti, Tony     2015

Now let’s get all the unique OCLC numbers from all the search results.

all_the_oclcs <- results[, unique(oclc)]
#>  [1] "1005106045" "1089176194" "949229431"  "1242682069" "1242707288"
#>  [6] "1244405806" "1104264768" "1242685020" "1242707684" "1244406814"
#> [11] "1104846312"

On to the worldcat_api_locations_by_oclc function! Since this function takes one OCLC number at a time, we need to use a looping-construct to run the function with all the OCLC numbers in all_the_oclcs. We’ll be using the pblapply function (from the great pbapply package) to do this because we get a useful progress bar with no extra effort.

holds <- pblapply(all_the_oclcs,

Since the pblapply function returns a list of data.tables (on for each OCLC), we’ll use data.table’s rbindlist function to one data.table containing all the results.

all_holdings_dt <- rbindlist(holds)
#>          oclc institution_identifier            institution_name copies
#>        <char>                 <char>                      <char> <char>
#> 1: 1005106045                    FEM        The Ferguson Library      1
#> 2: 1005106045                    YDX        YBP Library Services      1
#> 3: 1005106045                    DUQ Duquesne University Library      1

all_holdings_dt[, .(institution_name)]
#>                                            institution_name
#>                                                      <char>
#>    1:                                  The Ferguson Library
#>    2:                                  YBP Library Services
#>    3:                           Duquesne University Library
#>    4:                                    Centennial College
#>    5:                                  George Brown College
#>   ---
#> 1029:        Hochschule Mittweida (FH), Hochschulbibliothek
#> 1030:                                           Cyberlibris
#> 1031:                                           Cyberlibris
#> 1032:                                           Cyberlibris
#> 1033: Bibliothèque de l'Université du Québec à Trois-Riv...

There you have it! My textbook is held by 1033 distinct OCLC institutions!

Although the example above only searches the holding institutions of one specific book, the idiom is most helpful/interesting/cool when used for finding the holding institution of a entire class of materials.

For example, in a recent project for a curator at my institution, I used the Search API to find all search results for materials on a very specific topic, got all the holding institutions for each of the search results, and then aggregated the institutions (with this package’s dt_counts_and_percents function) to find the institutions holding the most items on this particular (very specific) topic.

Similar projects / code