Questions/comments regarding the configuration of Tree view

[Ansa211]

Tested on PDT 3.5 on Anša's instance, with the settings in /opt/kontext-ansa/configs/syntax-viewer.json:

• [ ] Why is PDT 3.0 configured so that the functor attribute is "filtered"? Why does removing the "filtered" element from syntax-viewer.json lead to an empty image of a tree when the plugin is used? What is the proper way of setting things up in such a way that functor could be shown in the node-detail section in the left part of the plugin output?

• [ ] What do the "id" and "name" settings in "syntax-viewer.json" do, and what are their possible values besides "default" and "Default"?

• [ ] Why are both id and word still present in the node-detail section even when I set the "filtered" option to "filtered": ["functor","id","word"],?

• [ ] When the option "Hide empty attributes" in the node-detail section (in the plugin pop-up) is ticked, the attribute e-parent with value <empty> should also be hidden. (For example, the node directly before the root of the tree has the empty value there.)

• [ ] Concerning the "attrRefs" setting: Each node has at most one eparent, so the current way of showing it as a two-level structure (with a 0 on the first level) is bit of an overkill (BUT the two-level structure would be needed if the attribute listed under "attrRefs" would be multi-value -- the question whether it is or is not multi-value can be easily distinguished by looking into the registry file); on the other hand, the second level is somewhat under-specified, with the identifier 0 corresponding to the technical id invented by KonText (it would be better to show the value actually present in the data, which is a relative value), and 1 corresponding to a concatenation of all the attributes of the e-parent according to the value specified in "attrRefs" in syntax-viewer.json - it would be better if each of the configured attributes would get a separate line in the output and would be preceded by the actual name of the attribute. Settings:

  "attrRefs": {
  "eparent": ["word","afun","functor"]
  },

  What I get: What I want to get:

  *eparent        +2  
       |--------------------------|
       | word          o          |
       | afun          AuxP       |
       | functor       PAT        |
       |--------------------------|

• [ ] Why cannot I use parent in the attrRefs setting in "syntax-viewer.json"? It leads to an "list index out of range" error. Is this related to the fact that parent is simultaneously the value of "parrentAttr"?

• [ ] On the other hand, it would be nice to separate multivalue attributes into a little table:

• [ ] It would be amazing if the user could select between two ways of ordering the attributes in the node-detail panel. One would be the current order (alphabetical), which makes it easy to find a specific attribute by its name. The other one would be the order in which the attributes are listed in the registry file, because there usually is some logic to it (such as listing "word" and the most salient attributes first, or grouping related attributes even if their names do not necessarily start with the same prefix).

• [ ] Related: #196

• [x] Why I cannot find the attribute pdtvallex in the node-detail section? It is not mentioned even once in the "syntax-viewer.json".

[Ansa211]

Ad "Why I cannot find the attribute pdtvallex...?": Attributes with the empty value (as determined by emptyValuePlaceholders in syntax-viewer.json are hidden from the node-detail section. To see the attribute, just choose some verb!

[Ansa211]

• [ ] Cryptic errors such as this one are not very helpful:

[Ansa211]

• [ ] I keep getting the error Failed to get the syntax tree due to limited KWIC context (too long sentence). on a sentence with ~20 words. I erased the cache, recompiled the corpus with settings

  MAXCONTEXT 500
  MAXDETAIL 500

  and restarted kontext-ansa. The issue persists nonetheless.

The query was this

[Ansa211]

• [ ] Is there a reason not to simply turn on the tree-view for any corpora with the parsing information by default? How much coding would be needed for such a universal setup? I suppose it's not enough to use a star in place of the corpus name in configs/syntax-viewer.json, is it? If the system was ready for something like that, it should be simple enough to formulate a general template that would display word forms as the node values and take the edges from the "parent" attribute. But that would mean that there would have to be a way of specifying precedence of the the settings in configs/syntax-viewer.json (something like "the first settings that match are applied", while the default settings would be placed at the end of the file).

[Ansa211]

• [x] We have over 100 UD 2.3 corpora; they will have exactly same settings of Tree View, so wildcards in configs/syntax-viewer.json would be much appreciated.
• [x] Since the new update, the left panel looks much worse than before, and the scroll bar does not work properly so it is not even possible to see all of the tree - this is what it looks like on my biggest screen:
• [x] The words "than before" refer to this (there are two bits of screenshots among the points above): it used to be that there was a table with attribute names in the left column and attribute values in the right column, now it's meaningless numbers in the left column and two-row tables in the right, first row is attribute name, second row is attribute value. BUT It we cannot get that view now, could we at least sort the table numerically and not alphabetically (currently the sorting is 1, 10, 11, ... 18, 2, 20, ...)?

[Ansa211]

• [ ] An idea for an extension: would it make sense to highlight the KWIC also in the Tree View? And would it be possible?
• [ ] A new type of error: when I use CQL search and search for <s> (beginning of sentence structural element), and then try to open the Tree View, I get error,Query evaluation error 5: Unexpected token, at position -1 near ';': missing elements, e.g. with this query I assume that in this case the KWIC is not part of any sentence and therefore the plugin cannot identify which sentence to show.