This project has moved and is read-only. For the latest updates, please go here.

Controlling the selection of terminals

Jan 27, 2008 at 2:21 AM
An Excel formula is a lot like an expression except there are cell references to consider. At one level a cell reference is an identifier so the reference to the cell in column A, row 1 can be represented as an IdentifierTerminal. However the cell can be $A1 or A$1 or $A$1. These different option can be expressed in in Irony using a statement such as:


Which provides a more accurate description and would result in better information in the node list. However, Irony seems to want to return A1. It parses A correctly as a COLUMNREFERENCE and finds 1 to be a ROWREFERENCE. It then finds that 1 is also a number and seems to take this in preference. As a result it cannot match the types COLUMN_REFERENCE + NUMBER with any available grammar definition. It then works out that A1 is also valid (as IDENTIFIER) and is the longest so is preferred.

How can I stop the process from trying to match 1 as a number when its already found the match with ROW_REFERENCE? I can't take NUMBER away because they are valid elsewhere in an expression.


Jan 27, 2008 at 4:54 AM
You should create custom terminal for recognition of cell references instead of trying to assemble it from parts in grammar rules. This makes sense, as cell reference is in fact a single entity, and it participates in formulas as a whole, not by parts. From your post in Issues section I see you already follow this path.
Jan 27, 2008 at 12:12 PM
Thanks for the reply.

I'm really using an Excel formula as a tool to learn the functionality available via Irony. Excel formulas have quirks so it's useful to me to see what issues I run into.

There are a two motivation for wanting to define "cell" via grammar definition statements. One is that being able to define the structure of a cell declaritively is great. The second is that while cell is a self contained entity at one level, it's not really. A reference to $A$1 will always indicate the upper left corner of a sheet. Each of the other combinations is really a specification of one cell relative to the "current" cell. If the current cell is B2 then A1 really means R-1C-1; $A1 means R-1C1 and finally A$1 means R1C-1 (Excel has an option to view formulas in the "R1C1" style instead of the more usual A1 format). When Excel saves a formula it does so using a stream of tokens and these tokens include separate elements for the row and column reference types which is useful. It seems that using a custom terminal to identify a whole cell reference means that the structure of a cell reference will be lost.

If I create a "CellTerminal" class, I'll be able to parse the cell and, presumably, record whatever I want about the cell reference, right?

Jan 29, 2008 at 6:19 AM
Sorry for not replying for some time; my answers are in different post in other discussion thread you started