BibUpload Admin Guide

1. Overview

BibUpload enables you to upload bibliographic data in MARCXML format into Invenio bibliographic database. It is also used internally by other Invenio modules as the sole entrance of metadata into the bibliographic databases.

Note that before uploading a MARCXML file, you may want to run provided /opt/invenio/bin/xmlmarclint on it in order to verify its correctness.

2. Configuring BibUpload

BibUpload takes a MARCXML file as its input. There is nothing to be configured for these files. If the files have to be coverted into MARCXML from some other format, structured or not, this is usually done beforehand via the BibConvert module.

Note that if you are using external system numbers for your records, such as when your records are being synchronized from an external system, then BibUpload knows about the 970 tag as the one containing external system number. (To change this 970 tag into something else, you would have to edit BibUpload config source file.)

Note that BibUpload also similarly knows about OAI identifiers, so it will refuse to insert the same OAI-harvested record twice, for example.

3. Running BibUpload

3.1 Inserting new records

Consider that you have an MARCXML file containing new records that is to be uploaded into the Invenio. (For example, it might have been produced by BibConvert.) To finish the upload, you would call the BibUpload script in the insert mode as follows:

$ bibupload -i file.xml



In the insert mode, all the records from the file will be treated as
new.  This means that they should not contain neither 001 tags
(holding record IDs) nor 970 tags (holding external system numbers).
BibUpload would refuse to upload records having these tags, in order
to prevent potential double uploading.  If your file does contain 001
or 970, then chances are that you want to update existing records, not
re-upload them as new, and so BibUpload will warn you about this and
will refuse to continue.


For example, to insert a new record, your file should look like this:


    <record>
        <datafield tag="100" ind1=" " ind2=" ">
            <subfield code="a">Doe, John</subfield>
        </datafield>
        <datafield tag="245" ind1=" " ind2=" ">
            <subfield code="a">On The Foo And Bar</subfield>
        </datafield>
    </record>

3.2 Inserting records into the Holding Pen

A special mode of BibUpload that is tigthly connected with BibEdit is the Holding Pen mode.

When you insert a record using the holding pen mode such as in the following example:

$ bibupload -o file.xml

the records are not actually integrated into the database, but are instead put into an intermediate space called holding pen, where authorized curators can review them, manipulate them and eventually approve them.

The holding pen is integrated with BibEdit.

3.3 Updating existing records

When you want to update existing records, with the new content from your input MARCXML file, then your input file should contain either tags 001 (holding record IDs) or tag 970 (holding external system numbers). BibUpload will try to match existing records via 001 and 970 and if it finds a record in the database that corresponds to a record from the file, it will update its content. Otherwise it will signal an error saying that it could not find the record-to-be-updated.

For example, to update a title of record #123 via correct mode, your input file should contain record ID in the 001 tag and the title in 245 tag as follows:

<record>
    <controlfield tag="001">123</controlfield>
    <datafield tag="245" ind1=" " ind2=" ">
        <subfield code="a">My Newly Updated Title</subfield>
    </datafield>
</record>

There are several updating modes:

-r, --replace Replace existing records by those from the XML
              MARC file.  The original content is wiped out
              and fully replaced.  Signals error if record
              is not found via matching record IDs or system
              numbers.
              Fields defined in Invenio config variable
              CFG_BIBUPLOAD_STRONG_TAGS are not replaced.

              Note also that `-r' can be combined with `-i'
              into an `-ir' option that would automatically
              either insert records as new if they are not
              found in the system, or correct existing
              records if they are found to exist.

-a, --append  Append fields from XML MARC file at the end of
              existing records.  The original content is
              enriched only.  Signals error if record is not
              found via matching record IDs or system
              numbers.

-c, --correct Correct fields of existing records by those
              from XML MARC file.  The original record
              content is modified only on those fields from
              the XML MARC file where both the tags and the
              indicators match: the original fields are
              removed and replaced by those from the XML
              MARC file.  Fields not present in XML MARC
              file are not changed (unlike the -r option).
              Fields with "provenance" subfields defined in
              'CFG_BIBUPLOAD_CONTROLLED_PROVENANCE_TAGS'
              are protected against deletion unless the
              input MARCXML contains a matching
              provenance value.
              Signals error if record is not found via
              matching record IDs or system numbers.

-d, --delete  Delete fields of existing records that are
              contained in the XML MARC file. The fields in
              the original record that are not present in
              the XML MARC file are preserved.
              This is incompatible with FFT (see below).

Note that if you are using the --replace mode, and you specify in the incoming MARCXML a 001 tag with a value representing a record ID that does not exist, bibupload will not create the record on-the-fly unless the --force parameter was also passed on the command line. This is done in order to avoid accidentally creating gaps in the database list of record identifiers. In fact, when you ask to --replace a non-existing record imposing a record ID with a value of, say, 1 000 000 and, subsequently, you --insert a new record, this will automatically receive an ID with the value 1 000 001.

If you combine the --pretend parameter with the above updating mode, you can actually test what would be executed without modifying the database or altering the system status.

3.4 Inserting and updating at the same time

Note that the insert/update modes can be combined together. For example, if you have a file that contains a mixture of new records with possibly some records to be updated, then you can run:

$ bibupload -i -r file.xml



In this case BibUpload will try to do an update (for records having
either 001 or 970 identifiers), or an insert (for the other ones).


3.6 Uploading fulltext files

The fulltext files can be uploaded and revised via a special FFT
("fulltext file transfer") tag with the following
semantic:


    FFT $a  ...  location of the docfile to upload (a filesystem path or a URL)
        $d  ...  docfile description (optional)
        $f  ...  format (optional; if not set, deduced from $a)
        $m  ...  new desired docfile name (optional; used for renaming files)
        $n  ...  docfile name (optional; if not set, deduced from $a)
        $o  ...  flag (repeatable subfield)
        $r  ...  restriction (optional, see below)
        $s  ...  set timestamp (optional, see below)
        $t  ...  docfile type (e.g. Main, Additional)
        $v  ...  version (used only with REVERT and DELETE-FILE, see below)
        $x  ...  url/path for an icon (optional)
        $z  ...  comment (optional)
        $w  ... MoreInfo modification of the document
        $p  ... MoreInfo modification of a current version of the document
        $b  ... MoreInfo modification of a current version and format of the document
        $u  ... MoreInfo modification of a format (of any version) of the document

For example, to upload a new fulltext file thesis.pdf associated to record ID 123:

<record>
    <controlfield tag="001">123</controlfield>
    <datafield tag="FFT" ind1=" " ind2=" ">
        <subfield code="a">/tmp/thesis.pdf</subfield>
        <subfield code="t">Main</subfield>
        <subfield code="d">
          This is the fulltext version of my thesis in the PDF format.
          Chapter 5 still needs some revision.
        </subfield>
    </datafield>
</record>

The FFT tag can be repetitive, so one can pass along another FFT tag instance containing a pointer to e.g. the thesis defence slides. The subfields of an FFT tag are non-repetitive.

When more than one FFT tag is specified for the same document (e.g. for adding more than one format at a time), if $t (docfile type), $m (new desired docfile name), $r (restriction), $v (version), $x (url/path for an icon), are specified, they should be identically specified for each single entry of FFT. E.g. if you want to specify an icon for a document with two formats (say .pdf and .doc), you’ll write two FFT tags, both containing the same $x subfield.

The bibupload process, when it encounters FFT tags, will automatically populate the fulltext storage space (/opt/invenio/var/data/files) and metadata record associated tables (bibrec_bibdoc, bibdoc) as appropriate. It will also enrich the 856 tags (URL tags) of the MARC metadata of the record in question with references to the latest versions of each file.

Note that for the $a and $x subfields, the filesystem paths must be absolute (e.g. /tmp/icon.gif is valid, while Destkop/icon.gif is not) and they must be readable by the user/group of the bibupload process that will handle the FFT.

The bibupload process supports the usual modes correct, append, replace, insert with a semantic that is somewhat similar to the semantic of the metadata upload:

Metadata Fulltext objects being uploaded MARC field instances characterized by tags (010-999) fulltext files characterized by unique file names (FFT $n) insert insert new record; must not exist insert new files; must not exist append append new tag instances for the given tag XXX, regardless of existing tag instances append new files, if filename (i.e. new format) not already present correct correct tag instances for the given tag XXX; delete existing ones and replace with given ones correct files with the given filename; add new revision or delete file; if the docname does not exist the file is added replace replace all tags, whatever XXX are replace all files, whatever filenames are delete delete all existing tag instances not supported

Note that in append and insert mode

$m

is ignored.

In order to rename a document just use the the correct mode specifing in the $n subfield the original docname that should be renamed and in $m the new name.

Special values can be assigned to the $t subfield.

Value

Meaning

PURGE

In order to purge previous file revisions (i.e. in order to keep only the latest file version), please use the correct mode with $n docname and $t PURGE as the special keyword.

DELETE

In order to delete all existing versions of a file, making it effectively hidden, please use the correct mode with $n docname and $t DELETE as the special keyword.

EXPUNGE

In order to expunge (i.e. remove completely, also from the filesystem) all existing versions of a file, making it effectively disappear, please use the correct mode with $n docname and $t EXPUNGE as the special keyword.

FIX-MARC

In order to synchronize MARC to the bibrec/bibdoc structure (e.g. after an update or a tweak in the database), please use the correct mode with $n docname and $t FIX-MARC as the special keyword.

FIX-ALL

In order to fix a record (i.e. put all its linked documents in a coherent state) and synchronize the MARC to the table, please use the correct mode with $n docname and $t FIX-ALL as the special keyword.

REVERT

In order to revert to a previous file revision (i.e. to create a new revision with the same content as some previous revision had), please use the correct mode with $n docname, $t REVERT as the special keyword and $v the number corresponding to the desired version.

DELETE-FILE

In order to delete a particular file added by mistake, please use the correct mode with $n docname, $t DELETE-FILE, specifing $v version and $f format. Note that this operation is not reversible. Note that if you don’t spcify a version, the last version will be used.

In order to preserve previous comments and descriptions when correcting, please use the KEEP-OLD-VALUE special keyword with the desired $d and $z subfield.

The $r subfield can contain a string that can be used to restrict the given document. The same value must be specified for all the format of a given document. By default the keyword will be used as the status parameter for the “viewrestrdoc” action, which can be used to give access right/restriction to desired user. e.g. if you set the keyword “thesis”, you can the connect the “thesisviewer” to the action “viewrestrdoc” with parameter “status” set to “thesis”. Then all the user which are linked with the “thesisviewer” role will be able to download the document. Instead any other user which are not considered as authors for the given record will not be allowed. Note, if you use the keyword “KEEP-OLD-VALUE” the previous restrictions if applicable will be kept.

More advanced document-level restriction is indeed possible. If the value contains in fact:

  • email: john.doe@example.org: then only the user having john.doe@example.org as email address will be authorized to access the given document.
  • group: example: then only users belonging to the local/external group example will be authorized to access the given document.
  • role: example: then only the users belonging to the WebAccess role example will be authorized to access the given document.
  • firerole: allow .../deny...: then only the users implicitly matched by the given firewall like role definition will be authorized to access the given document.
  • status: example: then only the users belonging to roles having an authorization for the WebAccess action viewrestrdoc with parameter status set to example will be authorized (that is exactly like setting $r to example).

Note, that authors (as defined in the record MARC) and superadmin are always authorized to access a document, no matter what is the given value of the status.

Some special flags might be set via FFT and associated with the current document by using the $o subfield. This feature is experimental. Currently only two flags are actively considered:

  • HIDDEN: used to specify that the file that is currently added (via revision or append) must be hidden, i.e. must not be visible to the world but only known by the system (e.g. to allow for fulltext indexing). This flag is permanently associated with the specific revision and format of the file being added.
  • PERFORM_HIDE_PREVIOUS: used to specify that, although the current file should be visible (unless the HIDDEN flag is also specified), any other previous revision of the document should receive the HIDDEN flag, and should thus be hidden to the world.

Note that each time bibupload is called on a record, the 8564 tags pointing to locally stored files are recreated on the basis of the full-text files connected to the record. Thus, if you whish to update some 8564 tag pointing to a locally managed file, the only way to perform this is through the FFT tag, not by editing 8564 directly.

The subfield $s of FFT can be used to set time stamp of the uploaded file to a given value, e.g. 2007-05-04 03:02:01. This is useful when uploading old files. When $s is not present, the current time will be used.

3.7 Obtaining feedbacks

Sometimes, to implement a particular workflow or policy in a digital repository, it might be nice to receive an automatic machine-friendly feedback that acknowledges the outcome of a bibupload execution. To this aim the --callback-url command line parameter can be used. This parameter expects a URL to be specified to which a `JSON <http://json.org/>`__-serialized response will POSTed.

Say, you have an external service reachable via the URL http://www.example.org/accept_feedback. If the argument:

--callback-url http://www.example.org/accept_feedback

is added to the usual bibupload call, at the end of the execution of the corresponding bibupload task, an HTTP POST request will be performed, if possible to the given URL, reporting the outcome of the bibupload execution as a JSON-serialized response with the following structure:

  • a JSON object with the following stringvalue mapping:
    • string: results – value: a JSON array whose values are all JSON objects with the following stringvalue mapping:
      • recid: an integer number, representing the described record identifier (-1 if no record identifier can be retrieved)
      • success: either true or false depending on the success of the elaboration of the corresponding MARCXML
      • error_message: a string containing a human-friendly description of the error that caused the MARCXML elaboration to fail (in case success was having false value)
      • marcxml: in case of success, this contains the final MARCXML representation of the record
      • url: in case of success, this contains the final URL where the detailde representation of the record can be fetched (i.e. its canonical URL)

For example, a possible JSON response posted to a specified URL can look like:

{
    "results": [
        {
            "recid": -1,
            "error_message": "ERROR: can not retrieve the record identifier",
            "success": false
        },
        {
            "recid": 1000,
            "error_message": "",
            "success": true,
            "marcxml": "1000...",
            "url": "http://www.example.org/record/1000"
        },
        ...
    ]
}

Note that, currently, in case the specified URL can not be reached at the time of the POST request, the whole bibupload task will fail.

If you use the same callback URL to receive the feedback from more than one bibupload request you might want to be able to correctly identify each bibupload call with the corresponding feedback. For this reason you can pass to the bibupload call an additional argument:

--nonce VALUE

where value can be any string you wish. Such string will be then added to the JSON structure, as in (supposing you specified --nonce 1234):

{
    "nonce": "1234",
    "results": [
        {
            "recid": -1,
            "error_message": "ERROR: can not retrieve the record identifier",
            "success": false
        },
        {
            "recid": 1000,
            "error_message": "",
            "success": true,
            "marcxml": "1000...",
            "url": "http://www.example.org/record/1000"
        },
        ...
    ]
}

3.8 Assigning additional information to documents and other entities

Some bits of meta-data should not be viewed by Invenio users directly and stored in the MARC format. This includes all types of non-standard data related to records and documents, for example flags realted to documents (sepcified inside a FFT tag) or bits of semantic information related to entities managed in Invenio. This type of data is usually machine generated and should be used by modules of Invenio internally.

Invenio provides a general mechanism allowing to store objects related to different entities of Invenio. This mechanism is called MoreInfo and resembles well-known more-info solutions. Every entity (document, version of a document, format of a particular version of a document, relation between documents) can be assigned a dictionary of arbitrary values. The dictionary is divided into namespaces, which allow to separate data from different modules and serving different purposes.

BibUpload, the only gateway to uploading data into the Invenio database, allows to populate MoreInfo structures. MoreInfo related to a given entity can be modified by providing a Pickle-serialised byte64 encoded Python object having following structure:

{
    "namespace": {
        "key": "value",
        "key2": "value2"
    }
}

For example the above dictionary should be uploaded as

KGRwMQpTJ25hbWVzcGFjZScKcDIKKGRwMwpTJ2tleTInCnA0ClMndmFsdWUyJwpwNQpzUydrZXknCnA2ClMndmFsdWUnCnA3CnNzLg==

Which is a base-64 encoded representation of the string

(dp0\nS'namespace'\np1\n(dp2\nS'key2'\np3\nS'value2'\np4\nsS'key'\np5\nS'value'\np6\nss.

Removing of data keys from a dictionary can happen by providing None value as a value. Empty namespaces are considered non-existent.

The string representation of modifications to the MoreInfo dictionary can be provided in several places, depending, to which object it should be attached. The most general upload method, the BDM tag has following semantic:

BDM $r  ... Identifier of a relation between documents (optional)
    $i  ... Identifier of a BibDoc (optional)
    $v  ... Version of a BibDoc (optional)
    $n  ... Name of a BibDoc (within a current record) (optional)
    $f  ... Format of a BibDoc (optional)
    $m  ... Serialised update to the MoreInfo dictionary

All (except $m) subfields are optional and allow to identify an entity to which MoreInfo should refer.

Besides the BDM tag, MoreInfo can be transfered using special subfields of FFT and BDR tags. The first one allows to modify MoreInfo of a newly uploaded document, the second of a relation. The additional subfields have following semantic:

FFT $w  ... MoreInfo modification of the document
    $p  ... MoreInfo modification of a current version of the document
    $s  ... MoreInfo modification of a current version and format of the document
    $u  ... MoreInfo modification of a format (of any version) of the document
BDR $m  ... MoreInfo modification of a relation between BibDocs

3.8.1 Uploading relations between documents

One of additional pieces of non-MARC data which can be uploaded to Invenio are relations between documents. Similarly to MoreInfos, relations are intended to be used by Invenio modules. The semantics of BDR field allowing to upload relations looks as follows

BDR $r  ... Identifier of the relation (optional, can be provided if modifying a known relation)

    $i  ... Identifier of the first document
    $n  ... Name of the first document (within the current record) (optional)
    $v  ... Version of the first document (optional)
    $f  ... Format of the first document (optional)

    $j  ... Identifier of the second document
    $o  ... Name of the second document (within the current record) (optional)
    $w  ... Version of the second document (optional)
    $g  ... Format of the second document (optional)

    $t  ... Type of the relation
    $m  ... Modification of the MoreInfo of the relation
    $d  ... Special field. if value=DELETE, relation is removed

Behaviour of the BDR tag in different upload modes:

insert, append Inserts new relation if necessary. Appends fields to the MoreInfo structure correct, replace Creates new relation if necessary, replaces the entire content of MoreInfo field.

3.8.2 Using temporary identifiers

In many cases, users want to upload large collections of documents using a single BibUpload task. The infrastructure described in the rest of this manual allows easy upload of multiple documents, but lacks facilities for relating them to each other. A sample use-case which can not be satisfied by simple usage of FFT tags is uploading a document and relating it to another which is either already in the database or is being uploaded within the same BibUpload task. BibUpload provides a mechanism of temporary identifiers which allows to serve scenarios similar to the aforementioned.

A temporary identifier is a string (unique in the context of a single MARC XML document), which replaces the document number or a version number. In the context of BibDoc manipulations (FFT, BDR and BDM tags), temporary identifiers can appear everywhere where version or numerical id are required. If a temporary identifier appears in a context of document already having an ID assigned, it will be interpreted as this already existent number. If a newly created document is assigned a temporary identifier, the newly generated numerical ID is assigned to the temporary id. In order to be recognised as a temporary identifier, a string has to begin with the TMP: prefix. The mechanism of temporary identifiers can not be used in the context of records, only with BibDocs.

A BibUpload input using temporary identifiers can look like this:

<collection xmlns="http://www.loc.gov/MARC21/slim">
  <record>
    <datafield tag="100" ind1=" " ind2=" ">
      <subfield code="a">This is a record of the publication</subfield>
    </datafield>
    <datafield tag="FFT" ind1=" " ind2=" ">
      <subfield code="a">http://somedomain.com/document.pdf</subfield>
      <subfield code="t">Main</subfield>
      <subfield code="n">docname</subfield>
      <subfield code="i">TMP:id_identifier1</subfield>
      <subfield code="v">TMP:ver_identifier1</subfield>
    </datafield>
  </record>

  <record>
    <datafield tag="100" ind1=" " ind2=" ">
      <subfield code="a">This is a record of a dataset extracted from the publication</subfield>
    </datafield>

    <datafield tag="FFT" ind1=" " ind2=" ">
      <subfield code="a">http://sample.com/dataset.data</subfield>
      <subfield code="t">Main</subfield>
      <subfield code="n">docname2</subfielxd>
      <subfield code="i">TMP:id_identifier2</subfield>
      <subfield code="v">TMP:ver_identifier2</subfield>
    </datafield>

    <datafield tag="BDR" ind1=" " ind2=" ">
      <subfield code="i">TMP:id_identifier1</subfield>
      <subfield code="v">TMP:ver_identifier1</subfield>
      <subfield code="j">TMP:id_identifier2</subfield>
      <subfield code="w">TMP:ver_identifier2</subfield>

      <subfield code="t">is_extracted_from</subfield>
    </datafield>
  </record>

</collection>

4. Batch Uploader

4.1 Web interface - Cataloguers

The batchuploader web interface can be used either to upload metadata files or documents. Opposed to daemon mode, actions will be executed only once.

The available upload history displays metadata and document uploads using the web interface, not daemon mode.

4.2 Web interface - Robots

If it is needed to use the batch upload function from within command line, this can be achieved with a curl call, like:

$ curl -F 'file=@localfile.xml' -F 'mode=-i' http://cds.cern.ch/batchuploader/robotupload [-F 'callback_url=http://...'] -A invenio_webupload



This service provides (client, file) checking to assure the records are put into a collection the client has rights to.
To configure this permissions, check CFG_BATCHUPLOADER_WEB_ROBOT_RIGHTS variable in the configuration file.
The allowed user agents can also be defined using the CFG_BATCHUPLOADER_WEB_ROBOT_AGENT variable.
Note that you can receive machine-friendly feedbacks from the corresponding
bibupload task that is launched by a given batchuploader request, by adding
the optional POST field callback_url with the same semantic of the --callback-url
command line parameter of bibupload (see the previous paragraph Obtaining feedbacks).

A second more RESTful interface is also available: it will suffice to append to the URL the specific mode (among "insert",
"append", "correct", "delete", "replace"), as in:

http://cds.cern.ch/batchuploader/robotupload/insert

The callback_url argument can be put in query part of the URL as in:

http://cds.cern.ch/batchuploader/robotupload/insert?callback_url=http://myhandler

In case the HTTP server that is going to receive the feedback at callback_url expect the request to be encoded in application/x-www-form-urlencoded rather than application/json (e.g. if the server is implemented directly in Oracle), you can further specify the special_treatment argument and set it to oracle. The feedback will then be further encoded into an application/x-www-form-urlencoded request, with a single form key called results, which will contain the final JSON data.

The MARCXML content should then be specified as the body of the request. With curl this can be implemented as in:

$ curl -T localfile.xml http://cds.cern.ch/batchuploader/robotupload/insert?callback_url=http://... -A invenio_webupload -H "Content-Type: application/marcxml+xml"

The nonce argument that can be passed to BibUpload as described in the previous paragraph can also be specified with both robotupload interfaces. E.g.:

$ curl -F 'file=@localfile.xml' -F 'nonce=1234' -F 'mode=-i' http://cds.cern.ch/batchuploader/robotupload -F 'callback_url=http://...' -A invenio_webupload

and

$ curl -T localfile.xml http://cds.cern.ch/batchuploader/robotupload/insert?nonce=1234&callback_url=http://... -A invenio_webupload -H "Content-Type: application/marcxml+xml"

4.2 Daemon mode

The batchuploader daemon mode is intended to be a bibsched task for document or metadata upload. The parent directory where the daemon will look for folders metadata and documents must be specified in the Invenio configuration file.

An example of how directories should be arranged, considering that Invenio was installed in folder /opt/invenio would be:

/opt/invenio/var/batchupload
       /opt/invenio/var/batchupload/documents
               /opt/invenio/var/batchupload/documents/append
               /opt/invenio/var/batchupload/documents/revise
       /opt/invenio/var/batchupload/metadata
               /opt/invenio/var/batchupload/metadata/append
               /opt/invenio/var/batchupload/metadata/correct
               /opt/invenio/var/batchupload/metadata/insert
               /opt/invenio/var/batchupload/metadata/replace

When running the batchuploader daemon there are two possible execution modes:

-m,   --metadata    Look for metadata files in folders insert, append, correct and replace.
                    All files are uploaded and then moved to the corresponding DONE folder.
-d,   --documents   Look for documents in folders append and revise. Uploaded files are then
                    moved to DONE folders if possible.

By default, the metadata mode is used.

An example invocation would be:

$ batchuploader --documents



It is possible to program the batch uploader to run periodically. Read the Howto-run guide to see how.