investigation of suitability of OpenFormula for word processors

i've tried to investigate what kind of (non-standard) formulas are currently implemented in Writer, and whether OpenFormula 1.3 would be suitable for word processors.

there are many open questions here that could benefit from comments by someone with OpenFormula experience, marked with "Q:".

AFAIK, OpenFormula has never been used in a word processor.

• documentation of existing formula support:
  • Writer: https://help.libreoffice.org/7.4/en-US/text/swriter/02/14020000.html?DbPAR=WRITER
  • Word:
    • https://support.microsoft.com/en-us/office/use-a-formula-in-a-word-or-outlook-table-cbd0596e-ea8a-485e-a35d-b2cb2c4f3e27
    • OOXML 17.16.3 Formulas and expressions
• formulas may occur on:
  • style:condition "is-true-formula" - not implemented
  • table:condition - content validation not implemented
  • table:expression - named expressions not implemented
  • table:formula - table cells
  • text:formula - fields
• potential issues for Writer:
  • sub tables (table:sub-table) - there is notation for nested cells
    ("SheetLocator"), so should be representable
  • nested tables - every table may have a name (table:name), so this can
    be used as the SheetName in OF expressions
  • built-in variables: there's a bunch of them:
    E (Euler's - missing in OF?), 7 document statistics ones,
    15 undocumented user-data ones "user_firstname" etc.
    • Q: do we need to add these?
  • formulas may refer to variables by name
    • type "variableName" = "string"
    • variables defined in 7.4.2 <text:variable-decls>, 7.4.7 <text:user-field-decls>
    • representable via quoting:
      SimpleNamedExpression ::= Identifier | '$$' (Identifier | SingleQuoted)
    • these must be interpreted as 5.11 "Named Expressions"
      • currently there is an obvious association with
        9.4.11 <table:named-expressions>
        table:named-expression / table:named-range
      • Q: unclear if associating NE with variable-decl requires
        changes to the text?
      • Q: unclear if requirement for case-insensitive names is an issue
      • evaluation must produce the current value of the variable
        (the same variable may be set to different values by multiple
        fields; evaluation depends on the position of the
        formula-carrying element relative to the set field)
  • formulas may refer to database addresses
    • string that contains DB name, table name, column name
      (dot-separated), may be quoted with []
    • also representable via quoting
    • current row is selected by database fields
      7.6.4 <text:database-next>
    • mostly the same concerns apply, except additionally:
      • Q: there is no declaration for these in the document like variable-decl?
        is there a requirement that every NE must be declared?
  • evaluation of name: try if name is a builtin or defined variable, otherwise try database
    Q: is this acceptable?
  • evaluation of field formulas may result in these types:
    bool, long, double, string, (error)
  • evaluation of cell formulas may result in these types:
    double, (error)
  • cell evaluation to numeric value: if it doesn't contain a formula or
    value, and starts with a field (preceded by whitespace) (one of which
    is 7.7.14 <text:table-formula> (deprecated)), the cell value is the
    field value, otherwise text parsed with number format
  • horizontally merged cells are addressed differently than in Calc:
    merge A1+B1, next cell in Writer/Word is B1, in Calc is C1
    Q: need to add a distinction text document vs. spreadsheet for
    merged cells in 9.2.1 Referencing Table Cells?
• conformance:
  • part4 2.3.1 B) "It shall conform to one of: (C16) OpenDocument Formula
    Small Group Evaluator, (C17) OpenDocument Formula Medium Group
    Evaluator or (C18) OpenDocument Formula Large Group Evaluator"
    • neither of these look suitable for text documents; maybe need to
      specify another one
  • Q: is this required by the part2 conformance?
    • it only requires syntactic conformance of formulas? there isn't
      an actual requirement of an evaluator conformant with part4?
• 3.4 Host-Defined Behaviors
  • Q: not sure what this implies - is it effectively just documentation?