I have a grammar that has a couple keywords that are optional. My MSchema wants a bool value - true if the keyword is there, false if the keyword is absent.

(not that I'm even close to my grammar generating M that is compatible with any schema, but that's the goal - though I begin to suspect any meaningful DSL syntax can't be translated directly to a relational schema using mgx.exe)

So I have this token:

@{Classification["Keyword"]}
token ReadOnly = "ReadOnly"
=> true;

And this syntax:

syntax SimpleProperty =
scope:Scope ident:Identity? ro:ReadOnly? type:Token name:Token EndStatement
=> Property { Name = name, Scope =scope, Identity=ident, ReadOnly = ro, Type = type };

Notice that ReadOnly is optional. This sort of renders into M, but not really.

If the keyword is there, the M output is 'true' including the ticks - which is not acceptible.

If the keyword is not there, the M output is null without any ticks, which is ok, but would require that this column allow nulls even though I really just want true/false.

I know I could probably solve this by having two syntax statements - one with the ReadOnly keyword and one without. But that's a really silly solution, because I have more than one optional keyword, so I'd end up with around 8 syntax blocks in my grammar, tons and tons of duplicated code just to handle this one issue.

Clearly this is a common scenario in languages. So I assume there must be a reasonable answer- probably something simple I'm missing.

Hi Rocky,
The short answer is that there isn't a way currently to eliminate quotes from values in the grammar output, but this is something that we are working on.

If you are trying to dump this into a database or the Repository, then you are going to need to fix up the M generated by MGX. There are a couple of options:
1. Create an import model with string fields and migrate the values to the 'real' model from there
2. Write a program with a regex to strip the quotes from the M file.

We are working on a sample for the second option.

BTW - It would be GREAT to get this entered as a bug in Connect so that others can vote on it.

---

Richard Orr

I blogged my code that does the fixup of the mgx output to solve both the boolean value issue, and another issue with single-item lists where the resulting M is entirely invalid.

http://www.lhotka.net/weblog/MCslaFixingUpOutputFromMgx.aspx

To answer myself (with some help from Justin Chase), a partial answer is this:

@{Classification["Keyword"]}
token ReadOnly = "ReadOnly";

syntax Boolean(e) =
e
=> true | empty => false;

And then to use this in my main syntax:

syntax SimpleProperty =
scope:Scope ro:Boolean(ReadOnly) type:Token name:Token EndStatement
=> Property { Name = name, Scope =scope, ReadOnly = ro, Type = type };


But I still get the single quotes around the value after using mgx.exe:

ReadOnly = 'false',

And those quotes cause issues with the M parser.

Hi Rocky,
The short answer is that there isn't a way currently to eliminate quotes from values in the grammar output, but this is something that we are working on.

If you are trying to dump this into a database or the Repository, then you are going to need to fix up the M generated by MGX. There are a couple of options:
1. Create an import model with string fields and migrate the values to the 'real' model from there
2. Write a program with a regex to strip the quotes from the M file.

We are working on a sample for the second option.

BTW - It would be GREAT to get this entered as a bug in Connect so that others can vote on it.

---

Richard Orr

I blogged my code that does the fixup of the mgx output to solve both the boolean value issue, and another issue with single-item lists where the resulting M is entirely invalid.

http://www.lhotka.net/weblog/MCslaFixingUpOutputFromMgx.aspx

I could have sworn we had fixed that bug, but I just tested your grammar against today's build and sure enough the quotes are there.

There obviously shouldn't be quotes around boolean literals.

Digging into it from our side.

DB