Epic: Syntax error highlighting

[allantokuda]

This is a toggleable editor mode, that when turned on,

• Highlights, in pale red, all LDS syntax errors (violations of the rules defined below). All items contributing to an error should be highlighted.
• When an item contributing to an error is clicked,
  • a message should appear explaining all syntax errors on that item.
  • it should visibly bolden in a brighter red, along with the other items that contribute to the same error.
  • If an item contributes to more than one syntax error, the first in priority order (as in the list below) will be the one that causes other elements to bolden.

A graph can be saved regardless of whether it has syntax errors.

Rules:

• Each entity, attribute, relationship, and link has an official name that is unique. (Ignore the relationship naming until that is a feature.)
• Identifiers:
  • Each entity has at least one identifier (in its attributes, or on one of its endpoints, there should be at least one bar).
  • There cannot be any cycles of identification dependency.
    • This happens if there is a barred link from A to B, and another from B to C, and a third from C back to A.
    • Both ends of a relationship cannot contribute to an identifier. This is a special case where A depends on B, and B in turn also depends on A.
    • A reflexive relationship cannot contribute to an identifier. This is a special case where A depends on itself.
  • There cannot be a bar and a chicken foot connected to an entity that has no other bars.
  • There cannot be a bar and no chicken foot connected to an entity that does have other bars.
  • There cannot be a bar opposite a chicken foot, unless the bar connects to an entity with no other bars.
  • No identifier can be a strict subset of another. If the secondary "squiggle"-type identifier is implemented, this becomes a possibility. On an entity, the items with squiggles cannot be a strict subset of the items with bars, or vice-versa.
• Labels:
  • A relationship has either zero or two labels.
  • All 1:1 relationships have labels.
  • All reflexive relationships have labels.
  • Between any pair of entities, there is at most one unlabeled relationship.
  • "To-be" relationships (labeled "be" at both ends):
    • They are all or nothing: you cannot label one end of a relationship "be" and the other end with anything other than "be".
    • They can only be on 1:1 relationships.
    • They cannot be on reflexive relationships.
    • Between any pair of entities, there is at most one "to-be" relationship.

These rules should be made into individual feature issues, once a an architecture is defined that will ultimately enable all of these to work.