Cannot get TBC import to work properly.

[UscholdM]

I was having problems with ontology import.  So I did some experiments.  In Protégé, I created a simple ontology that does nothing other than import another ontology, which is available on the web. I saved it, and when I opened it in Protégé, it behaved as expected. When I try to do so in TBC Free Edition 6.0.1 it does not work. 

I did further experimenting by downloading the imported ontology directly from the web and put the file in the TBCFE workspace. That also did not work. I then read the several paragraphs of TBC documention on working with imported ontologies. It seems like the base uri of an ontology plays a prominent role.  That seemed odd, since the ontology URI or ontology Version URI is what matters. Nevertheless, I tried manually inserting a base statement into the ontology to be imported setting it to the ontology URI and that also did not work. When I set the base URI to the Version IRI that was being explicitly imported, it worked.

The downloads page indicates that 6.0.1 is the latest stable free version. I asked my colleague to try it on his version, and it worked. I saw that he has the free edition, but his version was 6.1.1. How can that be? I asked other colleagues to try the same experiment and the importing was not working on their then-current version of TBCME, but when they installed the latest version it did work.

1. Attached is a very explicit set of things that I did, and that I sent to colleagues to report back on. It also includes screenshots of errors.
   
2. The tiny ontology that imports another ontology.
   

Punch line:

* Is there a free version of TBC that I can currently download that will import ontologies on the web that do not have a base statement?

* Is there any reason to give prominence to the base statement? It seems that what really matters is the IRI or VersionIRI of the imported ontology.

* Is there something else that could be causing the problem that adding a base statement solves by accident?

Thank you,

Michael

[Holger Knublauch]

Hi Mike,

On 19/11/2019 12:21, UscholdM wrote:

I was having problems with ontology import.  So I did some experiments.  In Protégé, I created a simple ontology that does nothing other than import another ontology, which is available on the web. I saved it, and when I opened it in Protégé, it behaved as expected. When I try to do so in TBC Free Edition 6.0.1 it does not work. 

I did further experimenting by downloading the imported ontology directly from the web and put the file in the TBCFE workspace. That also did not work. I then read the several paragraphs of TBC documention on working with imported ontologies. It seems like the base uri of an ontology plays a prominent role.  That seemed odd, since the ontology URI or ontology Version URI is what matters. Nevertheless, I tried manually inserting a base statement into the ontology to be imported setting it to the ontology URI and that also did not work. When I set the base URI to the Version IRI that was being explicitly imported, it worked.

The downloads page indicates that 6.0.1 is the latest stable free version. I asked my colleague to try it on his version, and it worked. I saw that he has the free edition, but his version was 6.1.1. How can that be? I asked other colleagues to try the same experiment and the importing was not working on their then-current version of TBCME, but when they installed the latest version it did work.

1. Attached is a very explicit set of things that I did, and that I sent to colleagues to report back on. It also includes screenshots of errors.
   
2. The tiny ontology that imports another ontology.
   

Punch line:

* Is there a free version of TBC that I can currently download that will import ontologies on the web that do not have a base statement?

Not that I know of.

* Is there any reason to give prominence to the base statement? It seems that what really matters is the IRI or VersionIRI of the imported ontology.

The reason why we don't use the owl:versionIRI is that this would require loading the graph first, just to detect the URI. With large files this is not feasible. Also, the web import needs to start somewhere. From the owl:imports statement we only have the physical URL, so we try to resolve the graph from there. If the versionIRI was different from the physical URI, how could a web import possibly work?

(In your example the versionIRI is identical to the physical location so I don't understand why this topic is relevant).

* Is there something else that could be causing the problem that adding a base statement solves by accident?

I installed 6.0.1 and can reproduce that it cannot directly import from the web. Maybe that particular version had a bug, or the server response (HTTP headers/mime type?) from this particular location is a problem for this version. It works fine in later versions, and I can also do Web imports for other URLs with 6.0.1 with no problem.

In any case, here is a work-around:

1) Right-click on some project and select Import RDF file from the Web

2) Fill in as shown

This produces a local file starting with

# baseURI: https://ontologies.semanticarts.com/o/gistTop9.0.0

which is our usual convention to recognize graph URIs. This line is saved at the beginning of each TTL file so that the system doesn't need to first scan the whole file for its owl:Ontology or owl:versionInfo or whatever. It improves start-up time and is the only scalable solution for very large files.

We generally recommend using local files over web imports as the latter may be unstable (server down, changing data etc).

HTH
Holger

[Michael F Uschold]

Thanks Holger, this is helpful.  A few things to follow up on:

>  I installed 6.0.1 and can reproduce that it cannot directly import from the web.

Even if the file is in on a local hard drive, the import does not work w/o a base statement.  Did you reproduce that also?

> I can also do Web imports for other URLs with 6.0.1 with no problem

So can I, but so far, I cannot do so unless there is a base statement.

Some versions of TBC seem to work w/o a base statement, but the current free version (6.0.1 ) does not.  So the emphasis on the base statement does not always seem to matter. Are there any plans to release an update to the free edition beyond 6.0.1?  Actually, somehow, my colleague does have a later version (6.1.1), but it is no longer available from your downloads page.

At any rate, thank you for the workaround.

Michael

On Wed, Nov 20, 2019 at 6:38 PM Michael F Uschold <usc...@gmail.com> wrote:

--
You received this message because you are subscribed to a topic in the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/topbraid-users/w1PtxWgtjTY/unsubscribe.
To unsubscribe from this group and all its topics, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/fee9ee92-fd93-fa98-0f6c-957b1b4dfd53%40topquadrant.com.

--

Sent from an input-challenged device.

--

Michael Uschold
   Senior Ontology Consultant, Semantic Arts
   http://www.semanticarts.com
   LinkedIn: www.linkedin.com/in/michaeluschold
   Skype, Twitter: UscholdM

[Irene Polikoff]

On Dec 1, 2019, at 7:55 PM, Michael F Uschold <usc...@gmail.com> wrote:

Thanks Holger, this is helpful.  A few things to follow up on:

>  I installed 6.0.1 and can reproduce that it cannot directly import from the web.

Even if the file is in on a local hard drive, the import does not work w/o a base statement.  Did you reproduce that also?

Correct. TBC needs to identify the graphs referred by the owl:imports statements. As Holger explained, from the owl:imports statement we get the URL of the referenced graph. First, TBC checks if a file matching a URL exists in the workspace. To be detected, the file must have this URL as its base URL. TBC has a File Registry that cross references each graph in its workspace with its URL. If not found in the workspace, then TBC will attempt to fetch it from the web using the specified URL. The web server response (HTTP headers/mime type) must be correct in order for this to work.

> I can also do Web imports for other URLs with 6.0.1 with no problem

So can I, but so far, I cannot do so unless there is a base statement.

Some versions of TBC seem to work w/o a base statement, but the current free version (6.0.1 ) does not.  So the emphasis on the base statement does not always seem to matter. Are there any plans to release an update to the free edition beyond 6.0.1?  Actually, somehow, my colleague does have a later version (6.1.1), but it is no longer available from your downloads page.

No plans. We currently only build TBC-ME edition.

<inbekjabdagdjhep.png>

2) Fill in as shown

<dclenjkiecnajbee.png>

This produces a local file starting with

# baseURI: https://ontologies.semanticarts.com/o/gistTop9.0.0

which is our usual convention to recognize graph URIs. This line is saved at the beginning of each TTL file so that the system doesn't need to first scan the whole file for its owl:Ontology or owl:versionInfo or whatever. It improves start-up time and is the only scalable solution for very large files.

We generally recommend using local files over web imports as the latter may be unstable (server down, changing data etc).

HTH
Holger

--
You received this message because you are subscribed to a topic in the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/topbraid-users/w1PtxWgtjTY/unsubscribe.
To unsubscribe from this group and all its topics, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/fee9ee92-fd93-fa98-0f6c-957b1b4dfd53%40topquadrant.com.

--

Sent from an input-challenged device.

--

Michael Uschold
   Senior Ontology Consultant, Semantic Arts
   http://www.semanticarts.com
   LinkedIn: www.linkedin.com/in/michaeluschold
   Skype, Twitter: UscholdM

--
You received this message because you are subscribed to the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CADfiEMMfu04Dssj5L0C845A2FCjhHsjbnM50_swpQ2OgOEw5iQ%40mail.gmail.com.

[Michael F Uschold]

Thanks for further clarification. I can work around the situation, but just out of curiosity... : 

> To be detected, the file must have this URL as its base URL.

This may be true for 6.0.1 of the Free Edition and may once have been true for other editions, but it does not seem to be true for the current releases. 

Michael

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/EAD098F9-5AE1-4CD5-9A0E-4114196D507B%40topquadrant.com.

[Holger Knublauch]

On 2/12/2019 13:33, Michael F Uschold wrote:

Thanks for further clarification. I can work around the situation, but just out of curiosity... : 

> To be detected, the file must have this URL as its base URL.

This may be true for 6.0.1 of the Free Edition and may once have been true for other editions, but it does not seem to be true for the current releases. 

I don't think the algorithm to match local files to remote URLs has been changed over the last few years. It does try to use a declared base URI, and if that fails it (for some file endings) attempts to load the file to find any declared owl:Ontology instance.

Holger

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CADfiEMNr4H-8r5kQW8Mk2sB4JiHGZPTor7R0%2B2ztENZ2cATdfQ%40mail.gmail.com.

[Michael F Uschold]

> I don't think the algorithm to match local files to remote URLs has been changed over the last few years.

I would not have thought so, which is why this behavior is curious.  It works on some releases but not others.

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/4be63ead-c06a-8578-7feb-b5ae633eb944%40topquadrant.com.

[Irene Polikoff]

I do not believe there is a release in which a local RDF file would be mapped to a URL identifying it without this file having a base URI. 

The exception to this rule are non RDF files such as spreadsheets. Their base URIs get derived from their location in the workspace, but this feature is not part of TBC Free, anyhow.

All of this is explained in details in Help. You can look it up in TBC by selecting Help or by going https://topbraidcomposer.org/html/NamedGraphsAndBaseURIs.htm. 

Of course, we can’t entirely rule out a possibility that some intermediate version has a bug that manifests itself under certain specific set of conditions. Having said this, the expected behavior is described in e-mails from Holger and I and in the help page.

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CADfiEMO%2BC%3D_h27q8-iC%2BwM9s3GA_PVUaEs7w58EwTEg84kbjjg%40mail.gmail.com.

[Michael F Uschold]

> I do not believe there is a release in which a local RDF file would be mapped to a URL identifying it without this file having a base URI. 

I dont think I ran that test. My experiments showed that if there is no base statement in an ontology, and if you want to load it from a URL on the web, it works for some versions of TBC but not others.  Perhap it was a bug, as you say. Or perhaps  in later versions, a base statement is added on the fly because TBC is smart enough to work out what it should be. TBC is generally extremely robust for imports, certainly compared to Protégé. 

M.

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/46BAB435-7791-4D5D-9D24-F5025C85CA04%40topquadrant.com.