tla / stemmatology Goto Github PK

The make_tradition script flags an error if the given input type doesn't exist, but a tradition is still created (and put in the database, if applicable.) This may also happen when an error is thrown. Either way the empty tradition should not be created.

If the TEI parser has to parse a file with namespaces more than once (or, presumably, a file without namespaces after a file with), it will break with an error like this:

XPath error : Invalid expression
//tei:tei:listWit/tei:tei:witness
         ^ at /opt/local/lib/perl5/site_perl/5.16.1/Text/Tradition/Parser/TEI.pm line 128.

so that graph export for RTL texts goes from R to L.

Integration with a phylogeny or other stemma generation package means that we will sometimes have unrooted stemmas. We need to give the user some way to assign a root (archetype) to an unrooted stemma.

The exception raising mechanism in Text::Tradition::Error tends to assume that whatever error it is asked to raise is a string. Moose has turned all of its exceptions into objects, which causes an exception if we try to treat it as a string when throwing an exception (confused yet?) We need to add a check for this.

Depending on the order returned by $collation->readings(), if you try to merge/collapse readings by relationship type and the relationship type has not been applied transitively (i.e. propagated) then the merge may or may not be complete. We need to do this recursively to be sure.

A <lem> tag in a TEI_PS file should result in readings with is_lemma set to true.

The Stemma object needs to be able correctly to parse a dot file with an undirected graph description, as generated from a Newick specification.

This needs to conform to what Classical Text Editor expects.

Need to implement the conversion of a Stemweb calculation into one or more Text::Tradition::Stemma objects tied to a particular tradition, as described in the API given here:
http://treeoftexts.arts.kuleuven.be/?p=58

A use case was submitted that makes use of the apparatus codicum, so that witStart tags and witEnd tags are present in the XML. These need to be parsed correctly.

In fixing #7 I found that the CSV export was breaking on a spurious decode_utf8() call. That should have been in a test.

The call to duplicate_readings can cause certain relationships to no longer be valid, and it will remove them. Right now this is done silently, but for UI purposes the relationship removal needs to be propagated.

The CSV export should allow tabs as well as commas for separation purposes.

Need finally to trace and zap the UTF-8 encoding bug in tradition names in the MySQL tables.

Tests start failing when we open a file handle to read from a UTF-8 character string. We need to stop doing that, by converting them to byte strings first.

It turns out that, when a check is made to prevent merge of readings that shouldn't be merged (i.e. as in 27e161b), a tight cycle is found in the equivalence graph. This breaks things.

All the IDP scripts need to be refactored around a central library.

Looks like a straightforward off-by-one error.

The Graph::Reader::Dot module fails to parse name tokens with characters outside the ASCII \w range unless the tokens are wrapped in double-quotes. Various hacks exist but this needs a real workaround.

The term 'lemma' is annoyingly overloaded, but...

When a reading is chosen as a lemma, its spelling and orthographic (and arguably punctuation) variants should have the same normal form as the lemma.

The CTE parser pretty badly needs some tests added to it...

...and our JSON parser ought to reflect this.

At the moment, any witness information is thrown away. This is unfortunate.

...because ' (a.c.)' fails XML Name validation.

At the moment it makes little sense to include a.c. witnesses in stemmatological reckoning. So when generating the collation table for those, we should exclude them. Related to tla/stemmaweb#29

The compress_readings method is failing when it shouldn't be. This is because of a naive means of assembling the original text in the sanity checking code.

It seems that when we check for transposition symmetry in the Analysis module, we've been comparing stringified versions of witness sets without sorting them first. Not sure how this ever reliably worked.

It is very common for scholars using CTE to note a transposition via 'post X transp.' in the apparatus. This has all kinds of pitfalls, but it might be nice if we make a first pass at solving it.

It should be possible to set a repetition relationship between a reading from X and a later reading from X (a.c.) - at the moment this is excluded.

Need to revamp handling of witStart and witEnd, to accord with what the XML is actually trying to represent. Current implementation is a misinterpretation.

We should not be calling ::Stemma for graph manipulation that doesn't concern an actual Stemma object. Thus need to refactor in order to avoid circular dependency on ::Stemma <-> ::StemmaUtil.

I realise that there hasn't been much development on this in the last decade, but I'm still interested as a perl user of 25 years and working on the text editing and analysis tool at menotag.ku.dk

I'm putting in CollateX JSON output but running into two problems: CollateX may have changed its format since this tool was updated, because (when using the JSON string):

Can't use an undefined value as an ARRAY reference at /usr/local/share/perl5/Text/Tradition/Parser/JSON.pm line 160, line 41.

Additionally, I'm using string input because Text::Tradition won't read the JSON file:

malformed JSON string, neither array, object, number, string or atom, at character offset 0 (before "(end of string)") at /usr/local/share/perl5/Text/Tradition/Parser/JSON.pm line 117.

Where the JSON file is perfectly fine.

It seems that the duplicate_reading function can, if used unwisely, lead to a bad graph. There needs to be a check to ensure that neither the new reading nor the old reading will get dissociated from the graph, which is to say, they both need to have at least one witness afterward.

Some users might want to collate a.c. readings in their manuscripts. If a witness is marked a.c., then the CollateX parser should treat it as a variant of the base witness and not as a new witness in its own right.