What are good PROV-N prefixes?

Most examples of PROV-N use example prefixes like:

prefix ex <http://example.com/>
prefix exg <http://example.org/government>

These example domains are explicitly reserved globally for all kinds of examples and training material, and deliberately do not have any content, advertisement or affiliations.

Assume you are writing the provenance of a student group exercise, should you be using the prefix/namespace ex and example.org to define agents/entities/relationship and your own attribute types?

You can keep using example.org, but ideally you should define your
own namespace based on a Web URL you “control” or “own”. This would make your Linked Data identifiers globally unique.

Students at The University of Manchester can publish their own home pages, and luckily most universities still provide a similar facility. Our students can search up themselves on https://personalpages.manchester.ac.uk/ and find for instance that they are http://personalpages.manchester.ac.uk/postgrad/alice.davidson/ which means they could make any sub-page under that directory.

Now the page does not need to exist – this blog is not about making Web pages – it is just important that it could exist at that address – that means it is that person’s identifier space.

So we could imagine that Alice made two files under her directory:

  • http://personalpages.manchester.ac.uk/postgrad/alice.davidson/terms
    Imagine this was a page defining new terms to be used in PROV
  • http://personalpages.manchester.ac.uk/postgrad/alice.davidson/groupExercise
    Imagine this page was the group exercise as text (which provenance is described by PROV), or the PROV-N file itself (self-describing)

In real life you would also have to deal with file extensions like .html or server configuration for content types, see Tim Berners-Lee’s Cool URIs don’t change for further considerations.

As above it is considered good practice to separate the prefix/namespace for new roles/attributes that you define vs specific agents/entities in one particular provenance trace. The idea being that the general terms could be reused in multiple provenance documents with the same meaning.

So let’s map in the above documents as namespaces in PROV-N:

prefix terms <http://personalpages.manchester.ac.uk/postgrad/alice.davidson/terms#>
prefix group <http://personalpages.manchester.ac.uk/postgrad/alice.davidson/groupExercise#>

The prefix string usually is a short name somewhat matching the address, it just needs be unique and consistent within the same PROV document. Other documents can freely map the same namespace to a different prefix.

A diligent reader might notice the # at the end of the namespace – this “fragment” is in Web documents used to indicate a subsection or heading within the same file.

This is so that terms:student expands to
<http://personalpages.manchester.ac.uk/postgrad/alice.davidson/terms#student>
rather than end with alice.davidson/termsstudent – even if we have not made the pages now we would not want to make a separate page for each word.

In some cases separate pages are desirable, in which case the namespace
ends in / to mark a directory – for instance s:Person becomes
http://schema.org/Person as schema.org have decided their term list is too big to keep in a single page.

Now we can use these prefixes for identifying
agents/entities/activities, as well as using our own attributes, roles
and types.

agent(group:Alice, [prov:type='prov:Person',
                    terms:favouriteDay="Monday"])

wasAssociatedWith(group:Alice, group:studying,
                  [prov:role='terms:Reader']) 

It is customary that attributes start with lower case after :, while
types/roles start with Capital – but this is just stylistic.

As for entity identifiers, your terms and roles should not have spaces or special characters in them as they must be combineable with the namespace to a valid URI, e.g. camel case favouriteDay or underscore favourite_day

It is possible in PROV-N (but not easily in other PROV syntaxes) to use freehand roles/types strings like prov:role="Writing carefully" – these are kind of anonymous and cannot be assumed to mean the same thing across multiple PROV documents.

In many cases there is no suitable URI yet, in which case using a temporary namespace like http://example.com/yourthing is perfectly valid as a working example, just be aware of the risk of other people having the same namespace idea!

Validating and visualising PROV

One of the advantages of W3C PROV having a common data model is that it can be serialized, or written out, in multiple file formats. The PROV family of W3C specifications describe mappings PROV-XML and PROV-O (which, being based on OWL2 itself has multiple serializations, for Linked Data including RDF formats Turtle and JSON-LD.

In addition to these standard approaches we also have PROV-JSON and PROV-JSONLD which could be well-suited for Web applications. All of these can in theory be mapped to each-other through the common PROV Data Model and the use of URIs as Linked Data global identifiers.

PROV-N

PROV also specifies its own language, PROV-N, a text-based file format that most closely represent the PROV Data Model. This representation is used by the PROV Primer to explain the PROV types (entity/agent/activity) and their relationships (e.g. wasAttributedTo). For example:

document
  prefix ex <http://example.com/>
  prefix s <http://schema.org/>
  
  entity(ex:regionList)
  entity(ex:dataset, [ prov:type='s:Dataset' ])
  entity(ex:composition)
  
  activity(ex:composing, [ prov:type='ex:Composing', 
     prov:label="Composing region and data" ])

  agent(ex:derek,
       [ prov:type='prov:Person', s:givenName="Derek", 
         s:email="derek@example.org" ])

  used(ex:composing, ex:dataset, 2011-11-16T16:00:00)
  used(ex:composing, ex:regionList, -)
  
  wasGeneratedBy(ex:composition, ex:composing,
                 2011-11-16T16:45:00)
  
  wasAssociatedWith(ex:composing, ex:derek, -)
  wasAttributedTo(ex:composition, ex:derek)

endDocument

The above PROV-N can be rendered as a diagram:

Let’s go through the PROV-N line by line:

  • prefix maps ex to the URI namespace starting with http://example.com/
    • PROV identifiers like ex:dataset1 can be expanded to a Linked Data global identifier http://example.com/dataset1 (which in an ideal world would describe or perhaps let you download the dataset)
    • External vocabularies like schema.org can be reused, e.g. the property 's:givenName' expands with the prefix s: to form the URI http://schema.org/givenName
    • prefix prov <http://www.w3.org/ns/prov#> is implicit, and is the internal namespace for PROV types and attributes.
    • Tip: It is possible to declare default <http://example.com/> after which ex:regionList can be shortened to regionList, however it is recommended to always use explicit prefixes to ease reuse and combination of PROV-N files.
  • entity(ex:regionList) declares the existence of an entity with that identifier. It can thereafter be used in relationships expecting an entity.
  • The entity ex:dataset is similarly declared, but also assigning a more specific type, using http://schema.org/Dataset from the external vocabulary.
  • The activity ex:composition is typed using an ad-hoc type ex:Composing from our own namespace, but also adds a string attribute to give a more descriptive label.
  • The agent identified as ex:derek is described with attributes from schema.org/Person
  • Relationships like used goes backwards in time
    • The activity ex:composing used the pre-exsting entity ex:dataset.
    • The usage happened on 16 Nov 2011 at 16:00 (given in ISO8601 date-time format)
  • The second use, this time of ex:regionList, has a placeholder - indicating that the required PROV-N argument for date-time is unknown.
  • The relationship wasGeneratedBy also points backwards in time, the new entity ex:composition was generated by the activity ex:composing some time later, at 16:45.
  • wasAssociatedWith indicates that our agent Derek took part in the ex:composing activity, with placeholder – as we don’t know when.
  • wasAttributedTo says Derek was (at least partially) invovled in generating the composition.

Some subtleties about PROV-N worth mentioning:

  • Like other Linked Data representations, PROV has an open world assumption, meaning that statements given may be a partial description of the actual provenance.
    • Additional statements carrying new knowledge can always be added, as long as they don’t break semantic constraints.
  • Statements can be listed in any order
    • It is convention to use a chronological partial order old…new so that the last lines in PROV represent the newest events.
  • Entities, Agents and Activities should be explicitly declared as such.
    • By convention declarations can be grouped together towards the top (as in above example)
    • Alternatively, a declaration can be listed just before the first reference to its identifier in other statements.
    • If a relationship is not showing up in PROV visualization, ensure it has the correct declaration.
  • Identifiers are globally unique according to the prefix mapping to URI namespaces.
    • Use of http://example.com/ namespaces is legal for examples/prototypes/training, but is at danger of collision if PROV graphs are combined.
    • To encourage Linked Data, as a minimum use a namespace leading to a human readable page, appending #
    • For instance PROV entities described within this blog post could use:
      prefix pp <https://practicalprovenance.wordpress.com/2020/11/13/validating-and-visualising-prov/#>
  • An entity can’t concurrently be an activity.
    • However an agent could concurrently be an entity or an activity

Two immediate questions arise when faced with this “new” syntax and language for provenance:

  1. How can we validate its syntax and the correct use of PROV types and arguments to PROV relations?
  2. How can we convert from/to PROV-N and file formats that are more accessible programmatically, such as PROV-JSONLD or PROV-O in Turtle?

PROV tooling

KCL’s openprovenance.org lists PROV supporting tools and libraries, including: ProvToolbox (Java), Prov Python, ProvJS. These libraries can be used by developers for generating or consuming PROV from within a programmatic environment like Jupyter Notebook or a data management application.

In addition there are graphical tools for PROV editing, validation, conversion and visualization described below:

PROV-N editor

The PROV-N Editor is an online text editor that provides syntax highlighting and autocomplete for PROV-N, and is useful for beginners new to PROV-N.

Screenshot of https://openprovenance.org/tools/editor/ as of 2020-11-13

Note that the starting example PROV-N aims to be somewhat complete, including the advanced use of nested bundle .…. endBundle block, //comments and deliberate invalid statements (shown in red).

We recommend using the PROV-N Editor starting with a simpler example, and to use copy-paste to save the PROV-N locally to a file, using a text editor like Visual Studio Code (which unfortunately do not have syntax highlighting for PROV-N):

PROV-N example from above edited in VSCodium

Note: The file extension for PROVN is .provn, but you may use .provn.txt to ensure it opens in a text editor. Do not edit PROV-N in a text processor like Microsoft Word, as its binary format .docx (actually a structured ZIP archive of XML files) is not parseable by PROV tools; in addition text processors may provide unhelpful assistance such as changing “quotes” to “curly quotes” which are not part of PROV-N syntax.

Validating PROV

Although the PROV-N editor does syntax highlighting and can detect glaring mistakes such as invalid file comments, it does not do deeper inspection to detect mistakes such as missing commas, mismatches parentheses, wrong or missing argument to PROV relations. You may also accidentally have added logically inconsistent statements, such as:

document
  prefix ex <http://example.com/back-to-the-future/>
  
  entity(ex:results)
  entity(ex:data)
  entity(ex:interviews)

  wasDerivedFrom(ex:results, ex:data)
  wasDerivedFrom(ex:data, ex:interviews)
  wasDerivedFrom(ex:interviews, ex:results)
endDocument

While the above “scruffy” PROV-N file is syntactically valid, and each of the statements are OK semantically, as a whole we seem to have added a semantic violation of causality; an entity can’t be generated from entities not yet existing. An attempt to draw the above as a diagram will show an endless loop of derivations:

To ensure your PROV-N is both syntactically valid and semantically consistent, it is best to use a PROV validator.

PROV Validator

The openprovenance.org PROV validator can support PROV-N; remember to tick the correct syntax, specially when pasting rather than uploading a file with the correct extension.

The checks performed by the PROV Validator mainly focus on semantic constraints such as correct typing and ensuring provenance goes backwards in time without any causality loops (e.g. you can’t be your own grandparent).

Unfortunately we have found that the PROV Validator service occasionally does not detect syntactic PROV-N errors, for instance if we delete the placeholder argument ,- from the wasGeneratedBy statement above it is silently accepted by this validator, even though the timestamp is required by PROV-N definition of used. If there are syntactic errors the user is not provided with line-numbers of where the error might be.

Therefore we also recommend using the PROV Toolbox command line tool to validate the PROV-N syntax before using the PROV Validator.

PROV Toolbox

The PROV Toolbox is a Java library for consuming and generating PROV, but it also includes a versatile command line tool that can do:

  • Validation
  • Conversion
  • Merging
  • Visualization
  • Generate PROV from templates

See PROV Toolbox tutorials for further information.

Installing PROV Toolbox

To use the command line tool, the PROV Toolbox must be installed locally on a desktop/laptop computer.

Installation requirements lists what is needed for compiling and development. For the command line tool we’ve found it is sufficient to have:

Binary packages of PROV Toolbox are included for Linux (RedHat/Centos, Debian/Ubuntu) and macOS although they are not always updated.

Note: Installing Java and PROV Toolbox in Windows users requires a series of steps that are detailed separately.

After installing or unzipping to a subdirectory you should be able to run its provconvert or bin/provconvert command:

(base) stain@biggie:~/software/ProvToolbox$ bin/provconvert -help
usage: provconvert [-allexpanded] [-bindformat <string>] [-bindings
       <file>] [-bindver <int>] [-builder] [-compare <file>] [-config]
       [-debug] [-flatten] [-formats] [-generator <string>] [-genorder]
       [-help] [-index] [-infile <file>] [-informat <string>] [-layout
       <string>] [-location <location>] [-log2prov <file>] [-merge <file>]
       [-namespaces <file>] [-outcompare <file>] [-outfile <file>]
       [-outformat <string>] [-package <package>] [-template <string>]
       [-templatebuilder <file>] [-title <string>] [-verbose] [-version]
 -allexpanded,--allexpanded                  In template expansion,
                                             generate term if all
                                             variables are bound.
 -bindformat,--bindformat <string>           specify the format of the
                                             bindings
 -bindings,--bindings <file>                 use given file as bindings
                                             for template expansion
                                             (template is provided as
                                             infile)
 -bindver,--bindver <int>                    bindings version
 -builder,--builder                          template builder
 -compare,--compare <file>                   compare with given file
 -config,--config                            get configuration
 -debug,--debug                              print debugging information
 -flatten,--flatten                          flatten all bundles in a
                                             single document (to used with
                                             -index option or -merge
                                             option)
 -formats,--formats                          list supported formats
 -generator,--generator <string>             graph generator
                                             N:n:first:seed:e1
 -genorder,--genorder                        In template expansion,
                                             generate order attribute. By
                                             default does not.
 -help,--help                                print this message
 -index,--index                              index all elements and edges
                                             of a document, merging them
                                             where appropriate
 -infile,--infile <file>                     use given file as input
 -informat,--informat <string>               specify the format of the
                                             input
 -layout,--layout <string>                   dot layout: circo, dot
                                             (default), fdp, neato, osage,
                                             sfdp, twopi
 -location,--location <location>             location of where the
                                             template resource is to be
                                             found at runtime
 -log2prov,--log2prov <file>                 fully qualified ClassName of
                                             initialiser in jar file
 -merge,--merge <file>                       merge all documents (listed
                                             in file argument) into a
                                             single document
 -namespaces,--namespaces <file>             use given file as declaration
                                             of prefix namespaces
 -outcompare,--outcompare <file>             output file for log of
                                             comparison
 -outfile,--outfile <file>                   use given file as output
 -outformat,--outformat <string>             specify the format of the
                                             output
 -package,--package <package>                package in which bindings
                                             bean class is generated
 -template,--template <string>               template name, used to create
                                             bindings bean class name
 -templatebuilder,--templatebuilder <file>   template builder
                                             configuration
 -title,--title <string>                     document title
 -verbose,--verbose                          be verbose
 -version,--version                          print the version information
                                             and exit

Here is an example of converting from provn to RDF Turtle.

(base) stain@biggie:~/software/ProvToolbox$ bin/provconvert -infile test.provn -outfile test.ttl

The example output is valid RDF and uses the same prefixes in a different notation. (This kind of output can be loaded in Triple stores like Jena Fuseki for further queries).

Note that as a UNIX-like tool, no output from provconvert means the conversion was successful. We can use provconvert for validation, even if we do not need the translated file. If the provn has syntax errors, this will be reported as:

(base) stain@biggie:~/software/ProvToolbox$ bin/provconvert -infile test.provn -outfile test.ttl
13:46:42,100  WARN Utility:35 - test.provn line 12:34 mismatched input ')' expecting ','

This tells us that in line 12, position 34, PROV-N expected an additional argument (the – placeholder) instead of the closing character ).

If you have installed Graphviz dot you can also make SVG or PNG images:

bin/provconvert -infile test.provn -outfile test.svg

Note that on Windows you would need to modify the PATH system variable for GraphViz to work, see installing PROV Toolbox for Windows.

PROV Store

PROV Store allows uploading of PROV documents, conversion and visualization. It is recommended to edit and validate PROV-N files with the methods listed above before uploading, as the PROV Store can be more picky on compliance with the PROV standards.

There seems to be a bug in email notifications not being sent when registering, so use the big “Register for free account” on https://openprovenance.org/store/account/signup/ which lets you straight in. Hack: For a second registration if email link has not been received, make a username like fred14 and add +14 to your email address: fred+14@example.org