Property talk:P212

There are several ISBN format (10 digits, 13 digits). Shouldn't we have several properties as well ? Or maybe we just want to have the more recent 13 digit format, but in that case, it should be indicated in the description. --Zolo (talk) 09:35, 7 March 2013 (UTC)Reply

  Done Let's start with ISBN-13. --Kolja21 (talk) 22:09, 7 March 2013 (UTC)Reply

On the long run we might need a property "ISBN-10", but for now we can use ISBN-13 Online Converter for older numbers. --Kolja21 (talk) 22:22, 7 March 2013 (UTC)Reply

How are we supposed to handle multiple translations of a book? The list of properties provides the example <The God Delusion> ISBN-13 <978-0-618-68000-9>. This makes perfect sense in English but Wikidata is multilingual and this ISBN is not the one of Pour en finir avec Dieu (French translation), Der Gotteswahn (German translation) or El espejismo de Dios (Spanish translation). If we include all of them, then it's impossible for readers to determine which ISBN corresponds to what. I'd be in favour of restricting this property to the ISBN of the original edition in the original language. Pichpich (talk) 19:28, 27 March 2013 (UTC)Reply

I think having an item for each edition (Wikidata talk:Notability#Books), though it would not solve the problem of the ISBN of the main item. I guess the main item shuold have the ISBN of the original edition, and we some Lua Module could probably find the ISBN for the translation. --Zolo (talk) 21:28, 27 March 2013 (UTC)Reply

okay, this is ISBN-13 now. but shouldn't there also be ISBN-10 as well? --Shisma (talk) 16:55, 28 March 2013 (UTC)Reply

ISBN-10 is not on Wikidata:Property proposal but I think it's uncontroversial. See also: Wikidata:Books task force. --Kolja21 (talk) 00:38, 29 March 2013 (UTC)Reply

I agree with Kolja21, we should have a page for each book, with is "national" ISBN. I agre we should also have a ISBN-10, just in case.

Still, we need to discuss relations between books. Please join the discussion here. --Aubrey (talk) 10:56, 29 March 2013 (UTC)Reply

I think that this property should be valid both ISBN-10 and ISBN-13 (changing the constraints), and that is should be renamed to "ISBN". The reasons for this is that if we want the ISBN to used in creating URIs to other websites, those websites typically handle both ISBN-10 and -13. This would also avoid confusionn in trying to do reverse lookups in Wikidata by ISBN, when that functionality arrives later. Maximilianklein (talk) 21:50, 10 September 2013 (UTC)Reply

@Kolja21:, @Zolo:, @Shisma:, @Aubrey:, we seem to be in agreement at least that we should start supporting ISBN-10, either by allowing this property to support both or to make another property. I would like to come to a consensus, and preferably to make this property support both 10 and 13, because I am ready to harvest Infobox Book. Any objections? Maximilianklein (talk) 19:20, 7 October 2013 (UTC)Reply

Using this property for both types of ISBN is fine with me, but then I do not know how other websites manage ISBN either. --Zolo (talk) 20:13, 7 October 2013 (UTC)Reply

This discussion was open for a week and I pinged all relevant parties for comment, of the one respondent, there was agreement. I believe this is consensus. @Akkakk: can you update the regex checker to support both 10 and 13 digit variants? 132.174.133.182 21:23, 14 October 2013 (UTC)Reply

I think it's better to create a new property ISBN 10, see Wikidata:Property proposal/Creative work#ISBN-10. It's easier to handle, avoids chaos and helps with the clean up. --Kolja21 (talk) 23:58, 14 October 2013 (UTC)Reply

Et voilà: ISBN-10 (P957) --Kolja21 (talk) 16:29, 15 October 2013 (UTC)Reply

@Kolja21: Can you please explain what you mean by "avoids chaos and helps with the clean up"? As for clean up, it seems like the regex here could be formatted so that only either a clean ISBN10 or ISBN13 could exist. As for avoids chaos, since ISBN10s can be converted to ISBN13s, it would be possible when having two distinct properties, that an item has two claims a ISBN-10 (P957) and a ISBN-13 (P212) and they would be really representing the same underlying edition. Maximilianklein (talk) 19:11, 17 October 2013 (UTC)Reply

With "chaos" I refer to the number of "format" violations on WD:Database reports (violations count: more than 1000). Since bots and humans edit WD, it seems at least helpful for the humans that there is the ten digit variant. I don't mind if ISBN-10 (P957) stays empty and a bot converts the numbers into ISBN-13. (BTW: Sorry, that I've not replied earlier, but the Notifications system has not yet been deployed to Wikidata.) --Kolja21 (talk) 21:44, 17 October 2013 (UTC)Reply

There should be no chaos at all: if both properties are defined, the ISBN-10 must match its conversion to ISBN 13 (i.e. only adding the prefix "978-" and recomputing the last check digit with the EAN/IUCC formula (multiply all 12 digits, including the prefix 978, by 1 and 3 alternatively, then compute their sum modulo 10) rather than the older formula (multiplying digits by 10,9,8,...,1 then compute their sum modulo 11, using "X" for the value 10). Both properties can only exist for books published up to 2006 (when only ISBN-10 was used). Beside that the grouping of digits in ISBN-13 has not changed, even if there's an additional leading prefix "978-".

However if books were published in 2007 or later, only the ISBN-13 should be present. And it's impossible to make the reverse conversion from ISBN-13 to ISBN-10 for any book published after 2007 using the new prefix "979". It is still very useful to have the ISBN-10, because this is the only code present in books printed before 2007, and still in most reference books (many of them have been published in the 19th or 20th century; the production of books in the 21th century is drammatically smaller, even if there are now some ebooks: people read less books today, they read articles in series or on the web as ebooks/PDF, or view videos; as well most reference publications today are not printed in books, but in series using ISSN, not ISBN). It is very likely that ISSN (still currently limited to 8-digits in two groups, including the last check digit computed modulo 11) will also be converted to the USN/UCC format with its own prefix, just because it will need extensions: the number of series (including webzines) has exploded dramatically in all countries; for now there are other competing standards for scientific series (such as ERIC), and ISSN remains used basically for printed newspapers and magazines, whose sales have also dramatially slowed down: there's a gap to fill for webzines and PDFs and to merge the efforts for scientific series, that have left the ISSN world (note that musical partitions have already abandonned the ISSN and adopted the ISBN-13; as well note that EAN/UCC has also already been extended to 14 digits for some classes of products, or more for fresh food to cary the variable price or weight or a date of production or conservation). Verdy p (talk) 20:53, 25 August 2020 (UTC)Reply

The ISBN numbers are different for every language version of a book, and I have worked on book templates on Swedish Wikipedia and we have the possibility to add ISBN number for:

• The original book
• The paperback version
• The audiobook on CD
• The audio book on MP3-CD
• The E-book
• Downloadable MP3 audiobook

And that is only the ISBN-numbers for the Swedish versions.

How do we solve this? I personally think that this is a database that should be filled with data, so should we make properties for Swedish paperback version, English paperback version, Norwegian paperback version, Swedish MP3-CD version, English MP3-CD version, and so on? /abbedabb^talk 13:30, 30 March 2013 (UTC)Reply

Right now we have no single edition books in Wikidata. We only have works. So strictly speaking the ISBN should be the ISBN of the first edition. We still have a long way to go before we can use Wikidata as a bibliography, see: Wikidata:Books task force. --Kolja21 (talk) 13:53, 30 March 2013 (UTC)Reply

Seconded and Wikidata:List of properties should be updated to make it clear that we don't want multiple ISBNs for a single work. Pichpich (talk) 16:53, 30 March 2013 (UTC)Reply

I've added the single value constraint to the property to indicate that there should only be one value. It will be possible to see how many items violate this rule here. 130.88.141.34 08:14, 7 June 2013 (UTC)Reply

I don't see why there shouldn't be multiple ISBN values. Why select a single one? If yes, which one? --  Docu  at 19:48, 7 June 2013 (UTC)Reply

• For the main book item the ISBN number is the number for the first edition.
• Where an item is for a particular edition then the ISBN number is for that edition.
• Where the main book item uses the edition number (P393) property to list other editions (rather than giving them separate items) then use the ISBN-13 (P212) property as a qualifier to give the ISBN number for that edition. Filceolaire (talk) 22:04, 9 September 2013 (UTC)Reply

In case of conference proceedings or journals, usually there are printed and electronic version of the same work, sometimes CD versions as well, with their own ISBN number. These are completely the same works, only the format (and the price) are different. Are you sure that we should create multiple bibliographic items for each formats? Wouldn't be better to define the ISBN numbers for the item with qualifiers, like print/electronic/e-book/cd etc. Samat (talk) 13:13, 14 December 2017 (UTC)Reply

@Samat: At the English Wikipedia, the practice is to SAYWHEREYOUGOTIT. If you found something in a specific edition, that is what you cite, not the first, nor the latest. You don't presume all formats or editions are equivalent: some national editions of books may be varied silently by their publishers to suit local sensibilities, politics, or tastes. One might also have a single consistent text with numerous applicable ISBNs simply for marketing/distribution reasons. We can't really second-guess all the variations. LeadSongDog (talk) 21:28, 9 August 2018 (UTC)Reply

the new format constraint pattern will assure that there are four dashes in the string. --Akkakk 01:10, 8 June 2013 (UTC)Reply

changed according to de:ISBN-Gruppennummer --Akkakk 15:34, 19 August 2013 (UTC)Reply

@Akkakk: Can you please change the format constraint to allow for the sequence "[\d]{3}\-[\d]{9}[\dxX]". That is three numbers a dash and then 10 numbers or a terminal X. It is a valid form, see this amazon book for instance but shows as an error in the constraint violation report. Maximilianklein (talk) 18:40, 5 November 2013 (UTC)Reply

@Maximilianklein: Amazon as a source? You're kidding. Open Library is full of Amazon records, one more horrible than the other. We should at least keep a minimum of scientific standards. I would delete all numbers that have been taken from there. --Kolja21 (talk) 20:43, 5 November 2013 (UTC)Reply

@Kolja21: I'm not saying that we should use Amazon or Open Library or any bad data as sources. I'm saying that 3digits-dash-10digits should be a valid P212 format. Maximilianklein (talk) 06:10, 6 November 2013 (UTC)Reply

@Maximilianklein: Of cause their are all kind of ways people and companies note the ISBN. But imho WD should use the official form. Why? Because "99921-58-10-7" provides the reader with the information that the ISBN is from Qatar. If you write the number without hyphens "9992158107" or like Amazon "978-9992158104" the ISBN still works but with disadvantages for the human reader. --Kolja21 (talk) 10:41, 6 November 2013 (UTC)Reply

Your format is clearly invalid for ISBN-13, as it includes the letter [xX] which is only valid in the last control character of ISBN-10 (computed modulo 11), while in ISO-13, the control character can only be a digit (computed modulo 10, with a different formula according to the EAN algorithm). OK the first group always has 3 digits in ISBN-13 (can only be 978 or 979 for now), it is followed by 9 significant digits (in 3 groups, whose length is variable, the 1st one identifying some languages, or a country or an international organization managing a local registry of publishers, then followed by the code of a publisher, whose length depend on the country code, then followed by the serial book number) then an hyphen, then the last check digit. See below for details (and a more complete dataset, summarized and matching official source isbn.org, an a few additions for the new "979" prefix)... Verdy p (talk) 05:54, 25 August 2020 (UTC)Reply

there is in my understanding no officially preferred presentation of the ISBN according to the standard. Therefore I changed the English label to not need to mention hyphenation. Maximilianklein (talk) 22:35, 24 October 2013 (UTC)Reply

That's sad. A computer don't care but human readers need the hyphens. I really think we should change the way of working. Right now WD imports to many data without control. If there are inconsistencies we lower the standards. We should use WD to improve the quality, instead of producing low quality duplicates. --Kolja21 (talk) 22:57, 24 October 2013 (UTC)Reply

I agree we should aim for high quality. Maybe getting there takes a while. Maybe you start with an unhyphenated number, and then later someone comes and hyphenates it. Maximilianklein (talk) 18:37, 5 November 2013 (UTC)Reply

Time has passed. Do we have a consensus on how ISBN-13s should be formatted? If we do, a bot can carry out the necessary changes. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:52, 27 August 2016 (UTC)Reply

@Pigsonthewing: The number to be stored is a 9-, 10- or bookland 13-digit integer, represented as strings. Any of the many forms of string in which it is acquired should be acceptable if it has the nine correct numerical digits, but the hyphens/dashes/spaces/whatever should be stripped out for storage. If a check-digit (vice "X") is provided, it should be validated and stored. If it is wrong, some form of exception handling is obviously needed. The correctness of the ISBN should be validated using queries to trustworthy repositories: surprisingly often an ISBN has multiple referents, rules notwithstanding, and sometimes a preliminary ISBN is used without the referent book being subsequently published. The ISBN-10 and ISBN-13 presentation format for easy human reading is a separate matter that may vary between users. I'd suggest just picking the presentation format called for in the standard, but others may disagree. I would think the bot role should be simply in putting those resultant formatted strings into citations in a way that is acceptable to the various projects' users. LeadSongDog (talk) 21:15, 9 August 2018 (UTC)Reply

The question was "Do we have a consensus..?" Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 23:01, 10 August 2018 (UTC)Reply

Note: my bot now inserts hyphens. So you can add data in digits-only format, its will be fixed automatically. — Ivan A. Krestinin (talk) 21:24, 9 August 2018 (UTC)Reply

Please see Wikidata:Request_a_query#Query_to_extract_Wikipedia_article_content_based_on_a_list_of_ISBNs.
--- Jura 06:58, 27 June 2018 (UTC)Reply

this page is now at Wikidata:Request_a_query/Archive/2018/06#Query_to_extract_Wikipedia_article_content_based_on_a_list_of_ISBNs --Shisma (talk) 14:40, 1 March 2020 (UTC)Reply

I added an ISBN-13 identifier for an instance of a book (Q60884133). The ISBN-13 statement shows a "Potential issues" warning: "conflicts-with constraint / An entity should not have a statement for ISBN-13 if it also has a statement for instance of with value book." Why is this happening? --Robert.Allen (talk) 04:39, 1 February 2019 (UTC)Reply

A separate item should be created for the version, edition or translation (Q3331189), and the ISBN should be added there. A book may have several editions with different ISBNs. —Tacsipacsi (talk) 21:31, 1 February 2019 (UTC)Reply

So this means that all books with only one edition have to have two Wikidata items? --Robert.Allen (talk) 02:51, 2 February 2019 (UTC)Reply

Yep, that’s the theory. To be honest, I don’t do this either, just ignore these warnings. If someone wants to be super precise, let them do this separation, but it really seems to be an overkill for one-edition books. The constraint, however, seems to be good for me: this is really a pontential issue, although not a certain one: many works have more than one edition, where the separation absolutely makes sense. —Tacsipacsi (talk) 14:05, 2 February 2019 (UTC)Reply

Thanks taking the time to respond. I understand much better now, and I think I completely agree with you. Best, --Robert.Allen (talk) 14:25, 3 February 2019 (UTC)Reply

Another thing to remember is that it is not just physical books (hard back, trade paperback etc) which produce different editions for books. Audiobooks and ebooks are also different editions of a book and can have ISBNs assigned to them. David Newton (talk) 14:40, 24 August 2019 (UTC)Reply

Hi, we have books, which have multiple ISBNs: e.g. Hazardní hráč (Q65144202). Could we remove "distinct values constraint" constraint? Skim (talk) 14:09, 8 July 2019 (UTC)Reply

No you don't, at least from the point of view of a "book" on Wikidata. ISBNs are assigned to individual editions of a book. If the book changes publisher that's potentially a new edition and thus new ISBN. If a book is translated that's certainly a new edition and thus a new ISBN. David Newton (talk) 14:42, 24 August 2019 (UTC)Reply

@David Newton: Hi David. This is not case when we have one edition from one publisher and second edition from second publisher. This is real situation on which one book has two ISBNs, because book was published by two publishing houses at time. See Hazardní hráč (Q65144202) and link in NKCR. This is common thing in some books in the Czech Republic. In citing template on Czech Wikipedia is parameter isbn and isbn2 Skim (talk) 20:09, 26 March 2020 (UTC)Reply

Well then that's two separate editions of the book, each edition having its own ISBN which exactly corresponds with what I said. David Newton (talk) 09:39, 27 March 2020 (UTC)Reply

@David Newton: no it's not, I had a case again today with Les manuscrits à peintures en France (Q109671685). It's one and only one edition (I have the said book in my hand right now, it's clearly one edition) with 2 editors and consequently 2 ISBN. @Skim: the constraint don't apply (anymore) if there is a qualifier (publisher (P123) in most cases). Cheers, VIGNERON en résidence (talk) 13:17, 22 November 2021 (UTC)Reply

@VIGNERON en résidence: Great, thanks. Skim (talk) 17:28, 22 November 2021 (UTC)Reply

Nope. Wrong again.

https://isbnsearch.org/isbn/2080121766

https://isbnsearch.org/isbn/2717719016

First one published by Flammarion as a paperback. Second one published by Flammarion as a hardback. Separate bindings and thus separate editions of the book, hence the separate ISBNs. To quote from the Wikipedia article on the subject, "An ISBN is assigned to each separate edition and variation (except reprintings) of a publication. For example, an e-book, a paperback and a hardcover edition of the same book will each have a different ISBN." David Newton (talk) 19:39, 21 December 2021 (UTC)Reply

@David Newton: I had on examplar in hand (which is obviously only one binding) where it's is clearly written on it the two ISBN for this specific edition, so I can assure you these ISBN Search results are wrong (and an other clue without seeing an actual examplar: 2-7177- is the ISBN prefix for the BnF, not for Flammarion). Cheers, VIGNERON en résidence (talk) 15:18, 10 January 2022 (UTC)Reply

Hi, The allowable item type constraint for ISBN-13 should include Q1143604 (proceedings) - as these are individual publications that sometimes have ISBNS (sometimes ISSNs and sometimes both). -- Phoebe (talk) 10:42, 15 August 2019 (UTC)Reply

Do you have an example? Perhaps you've mixed work and edition? --Kolja21 (talk) 15:49, 24 August 2019 (UTC)Reply

This property (and others that are about ISBN codes) should not use hyphens. Check the official ISBN documents: hyphens (or spaces!) are used only when ISBN have to be displayed in human-readable form. Using it in databases are not only unneeded, but may be a hindrance for end-users. The end-user should decide which form of presentation (raw, hyphens, spaces) is needed. Wostr (talk) 20:33, 17 January 2020 (UTC)Reply

@Wostr: this has been discussed many times. Removing hyphens is easy and trivial, adding hyphens is at best complicated and in some cases impossible. So, for the end-user it's far batter to put the hyphen, then they can decide what to do. Plus, I guess Wikidata fits the definition of a human-readable display. Cheers, VIGNERON (talk) 11:08, 24 August 2020 (UTC)Reply

adding hyphens is at best complicated and in some cases impossible — how so? Wostr (talk) 19:45, 24 August 2020 (UTC)Reply

Hi. Some books has bad ISBN on real book. And National library of the Czech Republic has this ISBN with addon "Bad ISBN" on book record. And there is another ISBN, which is right. But not published on real book. e.g. book Džihád - "80-86218-99-6 (váz.) * 80-86518-99-6 (chybné)". Could we somehow add these both ISBNs to book record and add some property for definition of correctness? Motive is simple. Some people don't know that ISBN on book is bad. e.g. my changes in. What do you mean? Skim (talk) 20:19, 26 March 2020 (UTC)Reply

User:Adamant.pwn and User:Haansn08 have changed the constraints a few days ago so that this can also be applied to book (Q571) and written work (Q47461344). Has there been any discussion about this? If not I'd revert the change. Editions have ISBNs, works do not. And there is no reason why the constraints of ISBN-10 and ISBN-13 should be different from each other. -- Discostu (talk) 20:50, 27 April 2020 (UTC)Reply

Where am I supposed to place ISBN if the book only has one edition? Should I create a separate entity for it? Adamant.pwn (talk) 21:02, 27 April 2020 (UTC)Reply

There has been no discussion I know of. Is there any guide for how to deal with the large number of items which are neither version, edition or translation (Q3331189) nor doctoral thesis (Q187685) but do have a ISBN-13 (P212) statement? It seems to me many are a written work (Q47461344) (including book (Q571)). -- Haansn08 (talk) 23:42, 27 April 2020 (UTC)Reply

There are many possibilities how to tackle this. I think this should be discussed with the WikiProject:Books   WikiProject Books has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. -- Discostu (talk) 10:41, 28 April 2020 (UTC)Reply

There is now a discussion about this at the WikiProject. -- Discostu (talk) 06:54, 13 May 2020 (UTC)Reply

Code to post at https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/queries/examples (or a subpage thereof) if you want

SELECT ?isbn ?statement WHERE {
   ?statement prov:wasDerivedFrom ?ref . 
   ?ref pr:P212 ?isbn .
   FILTER (CONTAINS(lcase(?isbn), lcase("$1")))
}

Replace $1 with the ISBN-13. Try it at https://query.wikidata.org/#SELECT%20%3Fisbn%20%3Fstatement%20WHERE%20%7B%0A%20%20%20%3Fstatement%20prov%3AwasDerivedFrom%20%3Fref%20.%20%0A%20%20%20%3Fref%20pr%3AP212%20%3Fisbn%20.%0A%20%20%20FILTER%20(CONTAINS(lcase(%3Fisbn)%2C%20lcase(%22%241%22)))%0A%7D ("$1"="%241")

I tried the above search (in the "Documentation" section I think) "haswbstatement:P212=978-2-26-702700-6" - https://www.wikidata.org/wiki/Special:Search?search=haswbstatement%3AP212%3D978-2-26-702700-6&fulltext=Search+for+%27978-2-26-702700-6%27&fulltext=Search&ns0=1&ns120=1, and it failed 2 of 2 times (one of the times I used an ISBN that I know is on a Wikidata page). Therefore, I made this query. --User123o987name (talk) 08:01, 28 May 2020 (UTC)Reply

Update: I don't know how to make a unified search for $3, where $3 is "ISBN-10 OR ISBN-13", aside from

SELECT ?isbn10 ?isbn13 ?statement WHERE {
   ?statement prov:wasDerivedFrom ?ref .
   ?ref pr:P957 ?isbn10 .
   FILTER (CONTAINS(lcase(?isbn10), lcase("$1")))
   ?statement prov:wasDerivedFrom ?ref . 
   ?ref pr:P212 ?isbn13 .
   FILTER (CONTAINS(lcase(?isbn13), lcase("$2")))
}

where you have to comment out lines 2-4 to search for ISBN-13 and comment out lines 5-7 if you want to search for ISBN-10. I tried the search thing for ISBN-10 -- https://www.wikidata.org/wiki/Property_talk:P957 ("haswbstatement:P957=0-521-22151-X") -- and it worked 2 out of 2 times, but here is a SPARQL version of the ISBN10 search: https://query.wikidata.org/#SELECT%20%3Fisbn10%20%3Fstatement%20WHERE%20%7B%0A%20%3Fstatement%20prov%3AwasDerivedFrom%20%3Fref%20.%0A%20%3Fref%20pr%3AP957%20%3Fisbn10%20.%0A%20FILTER%20(CONTAINS(lcase(%3Fisbn10)%2C%20lcase(%22%241%22)))%0A%7D

If you make a SPARQL query to search for $3, post it at https://www.wikidata.org/w/index.php?title=Wikidata:Book_sources&action=edit --User123o987name (talk) 08:42, 28 May 2020 (UTC)Reply

\d\d\d is the same as \d{3}, so could be update the regular expression to say:

97[89]-([0-57]-(\d-\d{7}|\d\d-\d{6}|\d{3}-\d{5}|\d{4}-\d{4}|\d{5}-\d{3}|\d{6}-\d\d|\d{7}-\d)|[89]\d-(\d-\d{6}|\d\d-\d{5}|\d{3}-\d{4}|\d{4}-\d{3}|\d{5}-\d\d|\d{6}-\d)|[69]\d\d-(\d-\d{5}|\d\d-\d{4}|\d{3}-\d{3}|\d{4}-\d\d|\d{5}-\d)|99[0-8]\d-\d-\d{4}|99[0-8]\d-\d\d-\d{3}|99[0-8]\d-\d{3}-\d\d|99[0-8]\d-\d{4}-\d|999\d\d-\d-\d{3}|999\d\d-\d\d-\d\d|999\d\d-\d{3}-\d)-\d

Instead of what it currently is:

97[89]-([0-57]-(\d-\d{7}|\d\d-\d{6}|\d\d\d-\d{5}|\d{4}-\d{4}|\d{5}-\d\d\d|\d{6}-\d\d|\d{7}-\d)|[89]\d-(\d-\d{6}|\d\d-\d{5}|\d\d\d-\d{4}|\d{4}-\d\d\d|\d{5}-\d\d|\d{6}-\d)|[69]\d\d-(\d-\d{5}|\d\d-\d{4}|\d\d\d-\d\d\d|\d{4}-\d\d|\d{5}-\d)|99[0-8]\d-\d-\d{4}|99[0-8]\d-\d\d-\d\d\d|99[0-8]\d-\d\d\d-\d\d|99[0-8]\d-\d{4}-\d|999\d\d-\d-\d\d\d|999\d\d-\d\d-\d\d|999\d\d-\d\d\d-\d)-\d|

Noticed this when working on https://www.wikidata.org/wiki/Wikidata_talk:Book_sources#SPARQL_query_to_improve --User123o987name (talk) 20:22, 29 May 2020 (UTC)Reply

@User123o987name: yes the number in the curly brackets is the number of character, \d{1} is \d, \d{2} is \d\d, \d{3} is \d\d\d, \d{4} is \d\d\d\d and so on. As \d{3} is short and easier to read I should be kept. That said, this regex has several other problems. Cheers, VIGNERON (talk) 12:21, 24 August 2020 (UTC)Reply

The "syntax clarification" qualifier for the regex currently reads

13 digits formatted in four groups separated by "-", where the 1st group must be "978" or "979".[...]The last group is a single check digit.

...and I may be too tired and setting myself up for ridicule here, but by my count both the regular expression as well as the typical format actually have five groups? Someone please double-check and clarify the clarification if I'm right. Matthias Winkelmann (talk) 03:36, 16 July 2020 (UTC)Reply

@Matthias Winkelmann: you're right, I changed it to "five". Cheers, VIGNERON (talk) 17:37, 20 August 2020 (UTC)Reply

I'm pinging @Verdy p: as there is other problems with this "clarification". At least the part « The 2nd group has 1 digit if it's "0" to "5" or "7", it has 3 digits if it starts by "6", it has 2 digits if it starts by "8", it has 2 to 5 digits if it starts by "9". » is only true for ISBN starting with 978 not 979 (the four groups existing right now being 979-8, 978-10, 979-11, 979-12 none fits that definition). Cheers, VIGNERON (talk) 12:33, 24 August 2020 (UTC)Reply

When I fixed it (partially) I used the list of violations and I reduced them a lot. There may be other cases (It's true that I did not check much the "979" prefix, as it is still not used a lot.

Anyway, the existing regexp only had (and still has) a single digit after "978" or "979", and it is the same 1-digit group that was used in the former ISBN format with 10 digits (when it was extended to 13 digits from the EAN standard, the legacy format had 9 digit (in 3 groups) + the last check digit isolaed in the last group; the new format used "978" for them, kept the 9 digits unchanged including their grouping, and only the last check digit was recomputed; then "979" was added in the EAN standard for use with ISBN, the 9 digits after are still in 3 groups but the 2nd group may be encoded differently to match the new needs; may be later another -digit prefix could be allocated in EAN for use in ISBN, as the EAN standard still has much place for new prefixes).

Note that after this change, I needed to let the property to be recomputed fully and it takes time for this Wiki to generate full reports. My intent was to reduce the number of exceptions after checking various examples (validated in actual books found in online libraries or citations).

When looking at all printed ISBN-13 numbers, I've still never seen "979-10-*" or "979-11-*" (and did not find them in existing Wikidata within contraint violation reports, but this may have changed now that bots have run). All samples I found were "978-1-*" (i.e. still 1 digit for the 2nd group when it is "0" to "5" or "7"). The 2nd group starting by "9" is still the most fragmented one (but it is very rare for the new prefixc "979").

The regexp must then evolve with what we find (the ISBN maintenance agency may allocate new blocks for editors in various countries, but may be using now more conservative rules, or could deallocate a large block partly to subdivide it; it's just like the allocation of IP addresses with variable-length prefixes: the rules can change if a block is underused and there are more needs for some publishers in other countries that can't use the current block and need new codes for their books; publishers may be small at one time but could become popular and start publishijng more than expected initially). Verdy p (talk) 19:11, 24 August 2020 (UTC)Reply

For clarity, here is the current format expanded which whitespaces and newlines (also indented) to show groups:

  (97[89] -
  ([0-57]      - (\d - \d{7} | \d\d - \d{6} | \d{3} - \d{5} | \d{4}-\d{4}|\d{5}-\d{3}| \d{6} - \d\d | \d{7} - \d)
  |8\d         - (\d - \d{6} | \d\d - \d{5} | \d{3} - \d{4} | \d{4}-\d{3}            | \d{5} - \d\d | \d{6} - \d)
  |6\d\d       - (\d - \d{5} | \d\d - \d{4} | \d{3} - \d{3}                          | \d{4} - \d\d | \d{5} - \d)
  |9([0-8]\d   - (\d - \d{5} | \d\d - \d{4} | \d{3} - \d{3}                          | \d{4} - \d\d | \d{5} - \d)
    |9([0-8]\d - (\d - \d{4} | \d\d - \d{3}                                          | \d{3} - \d\d | \d{4} - \d)
      |9\d\d   - (\d - \d{3}                                                         | \d\d  - \d\d | \d{3} - \d)
      )
    )
  ) - \d |

The last case for the 2nd group (starting by 9) if the most complex and may need to be refined further. as the violation report has run, I can check the remaining cases (as most violations that remain are for this 2nd group starting by 9.

Initially the format was even less flexible and had many more exceptions.

May be we could allow any grouping length, as long as there are 13 digits in 5 groups, and the 1st group has 3 digits (978 or 979), and the last group has only 1 digit. Note that the format still does not check the 13th digit in the last 5th group (even if it is computable from the 12 digits in the 4 leading groups). Verdy p (talk) 19:45, 24 August 2020 (UTC)Reply

I checked the current database and the rules are summarized like this ("/n" means the number of digits in the 2nd group, where "/0" means this is a reserved range currently not in use):

978 ISBN: [0-5]/1, 7/1, 65/2, 8/2, 9[0-4]/2, 6[0-4]/3, 9[5-8]/3, 99[0-8]/4, 999/5, 6[6-9]/0

979 ISBN: 8/1, 1[0-2]/2, 1[3-9]/0, [02-79]/0

So effectively the "979" prefix is used for now only for groups "979-10" to "979-13" and "979-8"; this means that the regexp should separate the "978" and "979" prefixes.

The current rule is too permissive for "978-66" to ""978-69" (currently still reserved but treated like "978-60" to "978-64").

It also creates a too long 2nd group for "978-65", and for "978-90" to "978-94". This can be fixed as:

97(8 - ( [0-57]              - (\d - \d{7} | \d\d - \d{6} | \d{3} - \d{5} | \d{4}-\d{4}|\d{5}-\d{3}| \d{6} - \d\d | \d{7} - \d)
       | (65 | 8\d | 9[0-4]) - (\d - \d{6} | \d\d - \d{5} | \d{3} - \d{4} | \d{4}-\d{3}            | \d{5} - \d\d | \d{6} - \d)
       | (6[0-4] | 9[5-8])\d - (\d - \d{5} | \d\d - \d{4} | \d{3} - \d{3}                          | \d{4} - \d\d | \d{5} - \d)
       | 99[0-8]\d           - (\d - \d{4} | \d\d - \d{3}                                          | \d{3} - \d\d | \d{4} - \d)
       | 999\d\d             - (\d - \d{3}                                                         | \d\d  - \d\d | \d{3} - \d)
       )
  |9 - ( 8                   - (\d - \d{7} | \d\d - \d{6} | \d{3} - \d{5} | \d{4}-\d{4}|\d{5}-\d{3}| \d{6} - \d\d | \d{7} - \d)
       | (1[0-2] | 6\d)      - (\d - \d{6} | \d\d - \d{5} | \d{3} - \d{4} | \d{4}-\d{3}            | \d{5} - \d\d | \d{6} - \d)
       )
  ) - \d |

However when compacting the string (removing whitespaces) this results in a too long string (498 characters) which was previously rejected by the Wikidata validator (now it is accepted but I think there's still a limitation in the Wikidata store for property values).

I could also make a longer regexp by validating the domain codes (by country, international orgs, or language) which also contain rules for publisher groups by country but this is much longer as these would require these additonal rules (it could be easily implemnted in Lua, if we could use a Lua validator function instead of a single long regexp, or if we could use multiple regexps), presented here in a more compact format, not regexps, here the "/n" indicate the length of the 3rd group for each publisher using the 2 groups at start of line, or "/0" if reserved):

978-0     en:  0-1/2,2-227/3,228-228/4,229-368/3,369-36/4,37-638/3,639-6397/4,6398-63/7,64-644/3,645-645/7,646-647/3,648-648/7,649-654/3,655-655/4,656-6/3,7-84/4,85-8/5,9-94/6,95-/7
978-1     en:  00/3,01-02/2,03-034/3,035-03/4,04-06/2,07-09/4,1-397/3,398-54/4,55-64/5,65-67/4,68-685/5,686-713/4,714-716/3,717-731/4,732-73/7,74-774/5,775-7753/7,7754-7763/5,7764-7764/7,7765-776/5,777-7776/7,7777-78/5,79-7/4,8-8004/5,8005-804/5,805-837/5,838-8384/7,8385-8671/5,8672-8675/4,8676-8697/5,8698-915/6,916-916505/7,916506-972/6,973-9877/4,9878-99114/6,99115-9911/7,9912-99898/6,99899-/7
978-2     fr:  0-1/2,2-34/3,35-3/5,4-48/3,49-494/6,495-495/3,496-4966/4,4967-496/5,497-6/3,7-83/4,84-8/5,9-9197/6,91980/5,91981-919942/6,919943-919968/7,919969-94/6,95-/7
978-3     de:  00-02/2,03-033/3,034-036/4,037-03/5,04-1/2,2-6/3,7-84/4,85-8/5,9-94/6,95-953/7,954-96/5,97-984/7,985-/5

978-4     JP:  0-1/2,2-6/3,7-84/4,85-8/5,9-94/6,95-/7
978-5     SU:  000-004/5,005-009/4,01-1/2,2-420/3,421-42/4,43/3,431-43/4,44-440/3,441-44/4,45-603/3,604-604/7,605-6/3,7-84/4,85-8/5,9-90/6,91/5,92/4,93-94/5,9500/7,9501-97/4,98/5,990/7,991-/4
978-600   IR:  0/2,1-4/3,5-8/4,9-9867/5,9868-992/4,993-995/3,996-/5
978-601   KZ:  0-1/2,2-6/3,7/4,8-84/5,85-/2
978-602   ID:  00-06/2,07-13/4,14/5,15-16/4,17-1/5,2-4/3,5-53/5,54-5/4,6-61/5,62-6/4,7-74/5,75-94/4,95-/5
978-603   SA:  00-04/2,05-4/2,5-7/3,8/4,/5
978-604   VN:  0-4/1,5-8/2,9-97/3,98-/4
978-605   TR:  00-02/2,03/3,04-0/2,1/3,2-23/4,24-3/3,4-5/4,6-74/5,75-7/4,8/5,/4
978-606   RO:  00-08/3,09-4/2,5-7/3,8-90/4,91/3,92-964/5,965-974/4,975-/3
978-607   MX:  0-3/2,4-74/3,75-94/4,95-/5
978-608   MK:  0/1,1/2,2-44/3,45-64/4,65-6/5,7-/1
978-609   LT:  0-3/2,4-7/3,8-94/4,95-/5
978-611   TH:  /0
978-612   PE:  0-2/2,3/3,4-44/4,45-4/5,5-514/4,515-/0
978-613   MU:  0-/1
978-614   LB:  0-3/2,4-7/3,8-94/4,95-/5
978-615   HU:  0/2,1-4/3,5-7/4,8/5,/0
978-616   TH:  0-1/2,2-6/3,7-8/4,/5
978-617   UA:  0-4/2,5-6/3,7-8/4,/5
978-618   GR:  0-1/2,2-4/3,5-7/4,8-/5
978-619   BG:  0-14/2,15-6/3,7-8/4,/5
978-620   MU:  0-/1
978-621   PH:  0-2/2,3/0,4-5/3,6-7/0,8/4,9-94/0,95-/5
978-622   IR:  0-08/2,09-1/0,2/3,3-5/0,6-7/4,8-94/0,95-/5
978-623   ID:  0/2,1/0,2-3/3,4-64/0,65-7/4,8/0,/5
978-624   LK:  0-04/2,05-1/0,2-24/3,25-4/0,5/4,6-94/0,95-/5
978-625   TR:  0/2,01-3/0,4-44/3,45-6/0,7/4,8-/0
978-65    BR:  0-01/2,02-24/0,25-2/3,3-302/3,303-4/0,5-512/4,513-54/0,55-5/4,6-7/0,8-81824/5,81825-85/0,86-8/5,9-90244/6,90245-98/0,/6
978-7     CN:  0/2,1-4/3,5-7/4,8/5,/6
978-80    CS:  0-1/2,2-6/3,7-84/4,85-8/5,9-998/6,/5
978-81    IN:  0-1/2,2-6/3,7-84/4,85-8/5,/6
978-82    NO:  0-1/2,2-68/3,6/6,7-8/4,9-98/5,/6
978-83    PL:  0-1/2,2-5/3,6/5,7-84/4,85-8/5,/6
978-84    ES:  0-11/2,12/6,13/4,14/3,15-1/5,2-6/3,7-84/4,85-8/5,9-91/4,92-923/6,924-92/5,93-94/6,95-96/5,97-/4
978-85    BR:  0-1/2,2-454/3,455-4552/6,4553-455/5,456-528/3,529-531/5,532-533/4,534-53/3,54-5402/5,5403/5,5404/6,5405-5408/5,540/6,541-543/5,544-547/4,548-54/5,55-5/4,6/5,7-84/4,85-8/5,9-924/6,925-944/5,945-95/4,96-97/2,98-/5
978-86    YU:  0-2/2,3-5/3,6-7/4,8/5,/6
978-87    DK:  0-2/2,3/0,4-64/3,65-6/0,7/4,8-84/0,85-94/5,95-96/0,97-/6
978-88    IT:  0-1/2,2-311/3,312-314/5,315-318/3,319-322/5,323-326/3,327-338/4,339-360/3,361-362/4,363-548/3,549-554/4,555-5/3,6-84/4,85-8/5,9-90/6,91-926/3,927-93/4,94-947/6,948-/5
978-89    KR:  0-24/2,25-54/3,55-84/4,85-94/5,95-96/6,97-98/5,/3
978-90    NL:  0-1/2,2-4/3,5-6/4,7/5,8-84/6,85-8/4,90/2,91-93/0,94/2,95-/0
978-91    SE:  0-1/1,2-4/2,5-64/3,65-6/0,7-81/4,82-84/0,85-94/5,95-96/0,97-/6
978-92    INT: 0-5/1,6-7/2,8/3,9-94/4,95-98/5,/6
978-93    IN:  0/2,1-4/3,5-7/4,8-94/5,95-/6
978-94    NL:  0-5/3,6-8/4,/5
978-950   AR:  0-4/2,5-8/3,9-98/4,/5
978-951   FI:  0-1/1,2-54/2,55-88/3,89-94/4,95-/5
978-952   FI:  0-1/2,2-4/3,5/4,6-65/2,66/4,67-6/5,7/4,8-94/2,95-98/4,/5
978-953   HR:  0/1,1-14/2,15-47/3,48-4/5,500/3,501-50/5,51-54/2,55-5/5,6-94/4,95-/5
978-954   BG:  0-28/2,2/4,3-7/3,8/4,9-92/5,93-/4
978-955   LK:  0-1/4,2-33/2,34-354/4,355-35/5,36-37/4,38/5,39-40/4,41-44/5,45-4/4,5-54/5,55-710/3,711-714/5,715-94/4,95-/5
978-956   CL:  0-08/2,0/5,1/2,2-5/3,6/4,7-/4
978-957   TW:  0-02/2,03-04/4,05-1/2,20/4,21-27/2,28-30/5,31-43/2,44-81/3,82-96/4,97-/5
978-958   CO:  0-48/2,49-50/3,51/4,52-53/5,54-55/4,56-5/5,6-7/3,8-94/4,95-/5
978-959   CU:  0-1/2,2-6/3,7-84/4,85-/5
978-960   GR:  0-1/2,2-65/3,66-68/4,6/3,7-84/4,85-92/5,93/2,94-97/4,98-/5
978-961   SI:  0-1/2,2-5/3,6-8/4,9-95/5,96-/0
978-962   HK:  0-1/2,2-6/3,7-84/4,85-86/5,87-8/4,/3
978-963   HU:  0-1/2,2-6/3,7-84/4,85-8/5,/4
978-964   IR:  0-14/2,15-24/3,25-2/4,3-54/3,55-8/4,9-96/5,97-98/3,/4
978-965   IL:  0-1/2,2-5/3,6/0,7/4,8/0,/5
978-966   UA:  0-12/2,13/3,14/2,15-16/4,17-1/3,2-278/4,279-28/3,2/4,3-6/3,7-8/4,90/5,91-94/3,95-97/5,98-/3
978-967   MS:  0/2,01-0/4,1/5,2-24/4,25-254/3,255-28/0,2/4,3-4/3,5/4,6-8/2,9-98/3,99-998/4,/5
978-968   MX:  01-3/2,4/3,5-7/4,8/3,/4
978-969   PK:  0-1/1,2-22/2,23/5,24-3/2,4-74/3,75-/4
978-970   MX:  0-5/2,6-8/3,9-90/4,91-96/5,97-/4
978-971   PH:  0-015/3,016-01/4,02/2,03-05/4,06-4/2,5-84/3,85-90/4,91-95/5,96/4,97-98/2,/4
978-972   PT:  0-1/1,2-54/2,55-7/3,8-94/4,95-/5
978-973   RO:  0/1,1-16/3,17-1/4,2-54/2,55-75/3,76-84/4,85-88/5,89-94/4,95-/5
978-974   TH:  0-1/2,2-6/3,7-84/4,85-8/5,9-94/5,95-/4
978-975   TR:  0-01/5,02-23/2,24/4,25-5/3,6-91/4,92-98/5,/3
978-976   CB:  0-3/1,4-5/2,6-7/3,8-94/4,95-/5
978-977   EG:  0-1/2,2-4/3,5-6/4,7-84/3,85-8/5,9-98/2,/3
978-978   NR:  0-1/3,2/4,3-7/5,8/4,/3
978-979   ID:  0/3,1-14/4,15-1/5,2/2,3/4,4-7/3,8-94/4,95-/5
978-980   VE:  0-1/2,2-5/3,6-/4
978-981   SG:  0-16/2,17/5,18-1/2,2/3,30/4,31-3/3,4-/4
978-982   QP:  0/2,1-6/3,7-8/2,9-97/4,98-/5
978-983   MS:  0-01/2,02-1/3,2-3/4,4-44/5,45-4/2,5-7/2,8/3,9-98/4,/5
978-984   BD:  0-3/2,4-7/3,8/4,/5
978-985   BL:  0-3/2,4-5/3,6-87/4,88-8/3,/5
978-986   TW:  0-11/2,12-53/3,54-7/4,8-/5
978-987   AR:  0/2,1/4,2/5,3-35/2,36-3/4,4-41/4,42-43/2,44/4,45-48/5,4/4,5-82/3,83-84/4,85-8/2,9-94/4,95-/5
978-988   HK:  0-11/2,12-1/5,2-73/3,74-76/5,77-7/5,8-96/4,97-/5
978-989   PT:  0-1/1,2-53/2,54/5,55-7/3,8-94/4,95-/5
978-9914  KE:  0-3/0,4-44/2,45-6/0,7-74/3,75-984/0,985-/4
978-9915  UR:  0-3/0,4-5/2,6-64/0,65-7/3,8-92/0,93-/4
978-9916  ET:  0/1,1-3/2,4/1,5/0,6-74/3,75-94/0,95-/4
978-9917  BO:  0/1,1-2/0,3-34/2,35-5/0,6/3,7-97/0,98-/4
978-9918  MT:  0/1,1/0,2/2,3-5/0,6-7/3,8-94/0,95-/4
978-9919  MN:  0-1/0,2-27/2,28-4/0,5/3,6-94/0,95-/4
978-9920  MA:  0-31/0,32-3/2,4-54/0,55-7/3,8/0,/4
978-9921  KW:  0/1,1-2/0,3/2,4-6/0,7-8/3,9-96/0,97-/4
978-9922  IQ:  0-1/0,2/2,3/0,6-7/3,8/0,/4
978-9923  JO:  0/1,1-4/2,5-6/0,7-8/3,9-96/0,97-/4
978-9924  KM:  0-2/0,3/2,4/0,5-64/3,65-8/0,/4
978-9925  CY:  0-2/1,3-54/2,55-734/3,735-/4
978-9926  BA:  0-1/1,2-3/2,4-7/3,8-/4
978-9927  QA:  0/2,1-3/3,4/4,5-/0
978-9928  AL:  0/2,1-3/3,4/4,5-/0
978-9929  GT:  0-3/1,4-54/2,55-7/3,8-/4
978-9930  CR:  0-4/2,5-93/3,94-/4
978-9931  DZ:  0-2/2,3-8/3,/4
978-9932  LO:  0-3/2,4-84/3,85-/4
978-9933  SY:  0/1,1-3/2,4-8/3,/4
978-9934  LV:  0/1,1-4/2,5-7/3,8-/4
978-9935  IS:  0/1,1-3/2,4-8/3,/4
978-9936  AF:  0-1/1,2-3/2,4-7/3,8-/4
978-9937  NE:  0-2/1,3-4/2,5-7/3,8-/4
978-9938  TN:  0-7/2,8-94/3,95-/4
978-9939  AM:  0-4/1,5-7/2,8/3,9-97/4,98-/2
978-9940  ME:  0-1/1,2-4/2,5-83/3,84-86/2,87-/4
978-9941  GE:  0/1,1-3/2,4-7/3,8/1,/4
978-9942  EC:  0-74/2,75-84/3,85-8/4,9-984/3,985-/4
978-9943  UZ:  0-2/2,3/3,4-974/4,975-/3
978-9944  TR:  0/4,1-4/3,5/4,6/2,7/3,8/2,/3
978-9945  DO:  0/2,01-07/3,08-3/2,4-56/3,57/2,58-7/3,80/2,81-84/3,85-/4
978-9946  KP:  0-1/1,2-3/2,4-8/3,/4
978-9947  DZ:  0-1/1,2-7/2,8-/3
978-9948  AE:  0-3/2,4-84/3,85-/4
978-9949  ET:  0-08/2,0/3,1-3/2,4-6/3,7-71/2,72-74/4,75-8/2,/4
978-9950  PS:  0-2/2,3-84/3,85-/4
978-9951  KV:  0-3/2,4-84/3,85-/4
978-9952  AZ:  0-1/1,2-3/2,4-7/3,8-/4
978-9953  LB:  0/1,1-3/2,4-5/3,6-8/2,9-92/4,93-96/2,97-/3
978-9954  MA:  0-1/1,2-3/2,4-7/3,8-98/4,/2
978-9955  LT:  0-3/2,4-92/3,93-/4
978-9956  CM:  0/1,1-3/2,4-8/3,/4
978-9957  JO:  0-3/2,4-64/3,65-67/2,68-6/3,7-84/2,85-87/4,88-/2
978-9958  BA:  0-01/2,02/3,03/4,04-08/3,09-0/4,1-18/2,19-1/4,2-4/2,5-8/3,/4
978-9959  LY:  0-1/1,2-7/2,8-94/3,95-96/4,97/3,98-/2
978-9960  SA:  0-5/2,6-8/3,/4
978-9961  DZ:  0-2/1,3-6/2,7-94/3,95-/4
978-9962  PA:  0-54/2,55/4,56-5/2,6-84/3,85-/4
978-9963  CY:  0-1/1,2-24/4,25-27/3,28-2/4,3-54/2,55-734/3,735-74/4,75-/4
978-9964  GH:  0-6/1,7-94/2,95-/3
978-9965  KZ:  0-3/2,4-8/3,/4
978-9966  KE:  0-13/3,14/2,15-1/4,2-6/2,7-74/4,75-820/3,821-824/4,825/3,826-828/4,829-95/3,96-/4
978-9967  KY:  0-3/2,4-8/3,/4
978-9968  CR:  0-4/2,5-93/3,94-/4
978-9970  UG:  0-3/2,4-8/3,/4
978-9971  SG:  0-5/1,6-8/2,9-98/3,/4
978-9972  PE:  0/2,1/1,2-24/3,25-2/4,3-5/2,6-8/3,/4
978-9973  TN:  0-05/2,06-08/3,0/4,1-6/2,7-96/3,97-/4
978-9974  UR:  0-2/1,3-54/2,55-74/3,75-87/4,88-90/3,91-94/2,95-/2
978-9975  MD:  0/1,1-2/3,3-3/4,4-44/4,45-8/2,9-94/3,95-/4
978-9976  TZ:  0-4/1,5-57/4,58/3,59-8/2,9-98/3,/4
978-9977  CR:  0-8/2,9-98/3,/4
978-9978  EC:  0-2/2,3/3,4-94/2,95-98/3,/4
978-9979  IS:  0-4/1,5-64/2,65/3,66-75/2,76-8/3,/4
978-9980  PG:  0-3/1,4-8/2,9-98/3,/4
978-9981  MA:  0/2,1-15/3,16-1/4,2-7/2,8-94/3,95-/4
978-9982  ZM:  0-7/2,8-98/3,/4
978-9983  GM:  0-7/0,8-94/2,95-98/3,/4
978-9984  LV:  0-4/2,5-8/3,/4
978-9985  ET:  0-4/1,5-7/2,8/3,/4
978-9986  LT:  0-3/2,4-8/3,9-93/4,94-96/3,97-/2
978-9987  TZ:  0-3/2,4-87/3,88-/4
978-9988  GH:  0-3/1,4-54/2,55-74/3,75-/4
978-9989  MK:  0/1,1/3,2/4,3-5/2,6-94/3,95-/4
978-99901 BH:  0-4/2,5-7/3,8-/2
978-99902 --:  /0
978-99903 MU:  0-1/1,2-8/2,/3
978-99904 CW:  0-5/1,6-8/2,/3
978-99905 BO:  0-3/1,4-7/2,8-/3
978-99906 KW:  0-2/1,3-5/2,6/3,7-8/2,9-94/2,95-/3
978-99908 MW:  0-0/1,1-8/2,/3
978-99909 MT:  0-3/1,4-94/2,95-/3
978-99910 SL:  0-2/1,3-8/2,/3
978-99911 LS:  0-5/2,6-/3
978-99912 BW:  0-3/1,4-5/3,6-8/2,/3
978-99913 AD:  0-2/1,3-35/2,36-5/0,6-604/3,605-/0
978-99914 INT: 0-4/1,5-6/2,7/1,8/2,/3
978-99915 MV:  0-4/1,5-7/2,8-/3
978-99916 NM:  0-2/1,3-6/2,7-/3
978-99917 BN:  0-2/1,3-88/2,89-/3
978-99918 FO:  0-3/1,4-7/2,8-/3
978-99919 BI:  0-2/1,3/3,4-7/2,8-/3
978-99920 AD:  0-4/1,5-8/2,9-/3
978-99921 QA:  0-1/1,2-6/2,7-7/3,8-8/1,9-/2
978-99922 GT:  0-3/1,4-6/2,7-/3
978-99923 SV:  0-1/1,2-7/2,8-/3
978-99924 NI:  0-1/1,2-7/2,8-/3
978-99925 PY:  0/1,1/2,2/3,3/1,4-7/2,8-/3
978-99926 HN:  0/1,1-5/2,6-8/3,87-8/2,/2
978-99927 AL:  0-2/1,3-5/2,6-/3
978-99928 GE:  0/1,1-7/2,8-/3
978-99929 MN:  0-4/1,5-7/2,8-/3
978-99930 AM:  0-4/1,5-7/2,8-/3
978-99931 SC:  0-4/1,5-7/2,8-/3
978-99932 MT:  0/1,1-5/2,6/3,7/1,8-/2
978-99933 NE:  0-2/1,3-5/2,6-/3
978-99934 DO:  0-1/1,2-7/2,8-/3
978-99935 HT:  0-2/1,3-5/2,6/3,7-8/1,/2
978-99936 BT:  0/1,1-5/2,6-/3
978-99937 MO:  0-1/1,2-5/2,6-/3
978-99938 SR:  0-1/1,2-5/2,6-8/3,/2
978-99939 GT:  0-5/1,6-8/2,/3
978-99940 GE:  0/1,1-6/2,7-/3
978-99941 AM:  0-2/1,3-7/2,8-/3
978-99942 SD:  0-4/1,5-7/2,8-/3
978-99943 AL:  0-2/1,3-5/2,6-/3
978-99944 ET:  0-4/1,5-7/2,8-/3
978-99945 NM:  0-5/1,6-8/2,/3
978-99946 NE:  0-2/1,3-5/2,6-/3
978-99947 TJ:  0-2/1,3-6/2,7-/3
978-99948 ER:  0-4/1,5-7/2,8-/3
978-99949 MU:  0-1/1,2-8/2,/3
978-99950 KM:  0-4/1,5-7/2,8-/3
978-99951 --:  /0
978-99952 MI:  0-4/1,5-7/2,8-/3
978-99953 PY:  0-2/1,3-7/2,8-93/3,94-/2
978-99954 BO:  0-2/1,3-6/2,7-87/3,88-/2
978-99955 SR:  0-1/1,2-5/2,6-7/3,8-/2
978-99956 AL:  0-5/2,6-85/3,86-/2
978-99957 MT:  0-1/1,2-7/2,8-94/3,95-/2
978-99958 BH:  0-4/1,5-93/2,94/3,95-/3
978-99959 LU:  0-2/1,3-5/2,6-/3
978-99960 MW:  0-06/0,07-0/3,1-94/2,95-/3
978-99961 SV:  0-2/1,3-36/3,37-8/2,/3
978-99962 MN:  0-4/1,5-7/2,8-/3
978-99963 KM:  0-4/2,5-91/3,92-/2
978-99964 NI:  0-1/1,2-7/2,8-/3
978-99965 MO:  0-2/1,3-35/3,36-62/2,63-/3
978-99966 KW:  0-2/1,3-6/2,7/3,8-96/2,97-/3
978-99967 PY:  0/1,1-5/2,6-/3
978-99968 BW:  0-3/1,4-5/3,6-8/2,/3
978-99969 OM:  0-4/1,5-7/2,8-/3
978-99970 HT:  0-4/1,5-8/2,/3
978-99971 MY:  0-3/1,4-84/2,85-/3
978-99972 FO:  0-4/1,5-8/2,/3
978-99973 MN:  0-3/1,4-7/2,8-/3
978-99974 BO:  0/1,1-25/2,26-3/3,4-63/2,64/3,65-7/2,8-/3
978-99975 TJ:  0-2/1,3/3,4-7/2,8-/3
978-99976 SR:  0-1/1,2-5/2,6-7/3,8/0,/3
978-99977 RW:  0-1/1,2-3/0,4-6/2,7/3,8-994/0,995-/3
978-99978 MN:  0-4/1,5-6/2,7-/3
978-99979 HN:  0-4/1,5-7/2,8-/3
978-99980 BT:  0/1,1-2/0,3-5/2,6-7/0,8-/3
978-99981 MO:  0-1/1,2-2/0,3-5/2,6-7/0,8-/3
978-99982 BN:  0/1,1-4/0,5-62/2,63-8/0,9-96/3,97-/0
978-99983 SV:  0/1,1-4/0,5-6/2,7-94/0,95-/3
978-99985 TJ:  0-54/0,55-5/2,6-94/0,95-/3

979-10    FR:  0-1/2,2-6/3,7-8/4,9-975/5,976-/6
979-11    KR:  0-24/2,25-54/3,55-84/4,85-94/5,95-/6
979-12    IT:  0-1/0,2/3,3-594/0,595-5/4,6-7/0,8-84/5,85-/0
979-65    BR:  0/5,1-4/0,8/5,9/0                     -- (?) see remark below
979-8     US:  0-5/0,6/4,7-984/0,985000/7,985001-/0  -- (?) see remark below

So how can we add support for longer regexps, or for multiple alternate regexps (or for a validation by a Lua function)? Verdy p (talk) 01:58, 25 August 2020 (UTC)Reply

Now the validation errors are reduced. There are still a few (about 200), some are effectively errors (such as an ISBN-10 indicated as value; note that ISBN-10 used a different check digit computed modulo 11 which could also be an "X", but ISBN-13 uses the EAN method and computes a modulo 10 result, so it is different: you can convert a valid ISB-10 to ISBN-13 by prepending "978-", but the last check digit must be recomputed; the reverse is also possible from ISB-13 to ISBN 10 only if the EAN-13 prefix is "978", not "979", used now by many publishers in France, South Korea, Italy and the United States, and possibly later by others, notably in Germany, the United Kingdom, Spain, Russia, China, India, or Brazil when they'll request new number blocks they can't find in the older "978" prefix which is almost full; it is also possible that prefixes other than 978 and 979 may be allocated to ISBN by the EAN agency; note that in the table above there are some assignments still not used, or reserved for extensions of ISBN like musical partitions, or possibly for online ebooks publishers, or if ISBN and ISSN are merged and share their codes in the same EAN space, or for legal publications like patents or international treaties, or historic books in public libaries, databases, documentations, notices, lyrics; note also that EAN-13 is already part of another standard using longer codes...). Verdy p (talk) 03:44, 25 August 2020 (UTC)Reply

My tests have shown that it validates all existing data, there remain about 100 individual items (down from tens of thousands), which only misplace the hyphens. The extended data above can help fix them (isolate the last check digit in the last group; determine the 2 first groups from the row header; determine the length of the 3rd group starting by the listed prefixes or ranges and indicated by the "/" but this is not checked by the regexp which expects 5 groups and checks only the length of the 1st, 2nd, and last 5th groups; the 4th group is the remaining digit(s) before the last check digit. I have also updated the ISBN-10 regexp (which is still used for referencing many books published in 2006 or before, and not showing themselves the ISBN-13, and not always found in book search engines with their new ISBN-13).

Note that I have found and temporarily added "979-65-[08]eeee-bb-k" is already used by a few publishers in Brazil, but still not reflected by the ISBN.org reference data (which may be late in its updates). I think however that these codes were converted from ISBN-10, incorrectly using the new prefix "979" instead of "978" (these are probably errors from these Brazilian publishers, but these books were still printed like this in 2007; theoretically they should be in "978-65-[08]eeee-bb-k" which is assigned to the Brazilian ISBN registration authority).

The situation seems to be different with US publishers now starting to use "979-8" ("979-8-6eee-bbbb-k" or "979-8-985000e-b-k"), apparently without error; this assignment to the US ISBN-RA is still not reflected in the ISBN.org reference data but visiblke in various online book catalogs and official deposit libraries in US or elsewhere. Verdy p (talk) 14:37, 25 August 2020 (UTC)Reply

Conservation status of New Zealand indigenous vascular plants, 2017 (Q76346849) has a 15 digit ISBN apparently for web based publications. The 15 digit ISBN has been plonked in the property P212, and of course fails the constraints and looks very ugly when using {{cite Q}} see e.g., en:Alternanthera denticulata. MargaretRDonald (talk) 05:00, 1 December 2020 (UTC)Reply

In some cases publishers assign wrong ISBN. Library catalogs have a special field "invalid ISBN" for this cases. We could add a qualifier instead. -- JakobVoss (talk) 18:21, 4 December 2020 (UTC)Reply

Yes, please add a method of marking the ISBN as correct, even if it is invalid. The English Wikipedia uses Template:Listed invalid ISBN for this purpose, and its citation templates use (()) markup. Templates on en.WP could be modified to detect Wikidata ISBNs marked as invalid and apply the appropriate markup. Jonesey95 (talk) 19:54, 10 December 2020 (UTC)Reply

@MargaretRDonald: there is no such thing as ISBN-15, this is apparently just a plain mistake from the publisher.

@JakobVoss, Jonesey95: I did this Q76346849#P212 (I vaguely remember this is the usual way to deal with these unusual and rare cases, I did this in the past but can't find the item again...). WDYT?

Cheers, VIGNERON (talk) 17:39, 25 January 2021 (UTC)Reply

Like most things at Wikidata, I can't understand any of the nomenclature used in that property. My understanding of the word "deprecated" is not compatible with the status of this ISBN, which is "it is invalid, but it is printed in the publication, so we should use it anyway." Deprecation usually means that something is subject to removal, which is not what we want. Jonesey95 (talk) 00:12, 26 January 2021 (UTC)Reply

Hi,

• I don't think we should change ISBN-13 in ISBN-15 because it's not valid
• In this case deprecation is not optimal, it sould apply if the source is wrong, but here it was printed.
• We could eventualy a new proprerty for "invalid ISBN"
• I think a qualifer is the best option, maybe something like "reason for a valid ISBN" : "printed value" ?

I have added a reference to Conservation status of New Zealand indigenous vascular plants, 2017 (Q76346849) because the deprecated un-sourced values should be deleted.

— eru ^{[Talk] [french wiki]} 07:20, 26 January 2021 (UTC)Reply

See Property_talk:P3962. --- Jura 12:14, 10 February 2022 (UTC)Reply

Video game in china has ISBN（[1]）. So I have added the video game (Q7889) in subject type constraint (Q21503250). Nostalgiacn (talk) 06:35, 29 December 2022 (UTC)Reply

An automatic formatter to insert the dashes would be great!-- Jackie Bensberg (talk) 20:11, 30 April 2023 (UTC)Reply

Unfortunately I don’t think it’s realistic: the dashes separate groups, and these groups are not necessarily deterministic. For example, I chose two books from my shelf, one of them is ISBN 978-963-05-8630-6 (published by Akadémiai Kiadó (Q555093)), the other one is ISBN 978-963-9715-43-1 (published by Animus Kiadó (Q12813588)). One can now that the last dash is always before the last digit (check digit); that the first dash in an ISBN-13 is after the 978 or 979; but the other two dashes cannot be placed without consulting huge databases. The second group is a one- to five-digit country code, while the third group is the publisher ISBN prefix. There are hundreds of country codes and probably at least tens of thousands of publisher prefixes. —Tacsipacsi (talk) 22:54, 30 April 2023 (UTC)Reply

I use ISBN Converter. Your example 9789630586306:

• Converted ISBN 963-05-8630-4
• Converted ISBN 978-963-05-8630-6

--Kolja21 (talk) 23:48, 30 April 2023 (UTC)Reply

I've just noticed, that DeltaBot (talk • contribs • logs) automatically fixes ISBNs.--Jackie Bensberg (talk) 08:09, 1 May 2023 (UTC)Reply

I'm now also running a script periodically to fix issues that can be automatically fixed (see edit groups). It relies on the isbn3 JS library, which downloads the ISBN groups data from https://www.isbn-international.org Maxlath (talk) 12:28, 13 September 2024 (UTC)Reply

There are currently 813 violation of the "unique value" constraint, but the people that introduced them might not be aware of it, so I wrote a script to try to identify and notify them, in the hope they can fix those issues. Here is the resulting report per user: P212 unique value constraint violations by user.

Unfortunately, I'm limited in how many people I can notify, so here is a first batch: @Vojtěch Dostál, ShibaSalmon, John P. Sadowski (NIOSH), MonicaMu, KrzysztofPoplawski, Rdmpage:

-- Maxlath (talk) 13:36, 7 October 2024 (UTC)Reply

Next would be @Raliedir, Iamcarbon, Emijrpbot, Daniel Baránek, Harej: Maxlath (talk) 13:55, 7 October 2024 (UTC)Reply

Thanks for the ping. In my case, the database I import from routinely assigns conference abstracts/papers and book chapters the ISBN of the broader work containing them. In these cases I give the statements deprecated rank with qualifier reason for deprecated rank (P2241): identifier refers to broader entity (Q103875254) when I find them. I've made the changes for this batch. John P. Sadowski (NIOSH) (talk) 23:46, 12 October 2024 (UTC)Reply

The URL http://classify.oclc.org/classify2/Classify?isbn=$1 doesn't seem to be active anymore. Solstag (talk) 16:09, 5 November 2024 (UTC)Reply