[MySQL (Oracle one)] lvalueVariable rule, why the "DEFAULT_SYMBOL dotIdentifier" case?

[OnMYLai]

I've been dwelling through each statement rules in the grammar, when I run into the SET statement's, I saw this lvalueVariable rule which is basically used in case SET lvalueVariable = someExpr, can parse sth like DEFAULT .someVar.
I've looked through and through the MySQL SET Statement doc and found no sample of SET DEFAULT .someVar = someExpr, I've also test running a set query like this on my mysql db to no avail.
Is there some usecase of this lvalueVariable case that I'm not aware of?

[kaby76]

Yes, for the input SET DEFAULT .someVar = someExpr;, the parser accepts the input and produces this parse tree (via Trash Toolkit, which places off-channel content in the tree).

$ trparse -i 'SET DEFAULT .someVar = someExpr;' | trtree
CSharp 0 string success 0.0995957
queries
├── query
│   ├── simpleStatement
│   │   └── setStatement
│   │       ├── SET_SYMBOL
│   │       │   └── "SET"
│   │       └── startOptionValueList
│   │           ├── optionValueNoOptionType
│   │           │   ├── lvalueVariable
│   │           │   │   ├── Attribute WHITESPACE Value ' ' chnl:HIDDEN
│   │           │   │   ├── DEFAULT_SYMBOL
│   │           │   │   │   └── "DEFAULT"
│   │           │   │   └── dotIdentifier
│   │           │   │       ├── Attribute WHITESPACE Value ' ' chnl:HIDDEN
│   │           │   │       ├── DOT_SYMBOL
│   │           │   │       │   └── "."
│   │           │   │       └── identifier
│   │           │   │           └── pureIdentifier
│   │           │   │               └── IDENTIFIER
│   │           │   │                   └── "someVar"
│   │           │   ├── equal
│   │           │   │   ├── Attribute WHITESPACE Value ' ' chnl:HIDDEN
│   │           │   │   └── EQUAL_OPERATOR
│   │           │   │       └── "="
│   │           │   └── setExprOrDefault
│   │           │       └── expr
│   │           │           └── boolPri
│   │           │               └── predicate
│   │           │                   └── bitExpr
│   │           │                       └── simpleExpr
│   │           │                           └── columnRef
│   │           │                               └── fieldIdentifier
│   │           │                                   └── qualifiedIdentifier
│   │           │                                       └── identifier
│   │           │                                           └── pureIdentifier
│   │           │                                               ├── Attribute WHITESPACE Value ' ' chnl:HIDDEN
│   │           │                                               └── IDENTIFIER
│   │           │                                                   └── "someExpr"
│   │           └── optionValueListContinued
│   └── SEMICOLON_SYMBOL
│       └── ";"
└── EOF
    └── ""

As with many grammars, sometimes the grammar will accept something that should be illegal. This is because there needs to be a semantic analyzer. It says so in the grammar for lvalueVariable:

grammars-v4/sql/mysql/Oracle/MySQLParser.g4

Line 3211 in 63d037b

// Check in semantic phase that the first id is not global/local/session/default.

One could add type-checking to the grammar through the actions and semantic predicates, but generally, the practice is to place that in a phase after parsing.

[OnMYLai]

I thought that comment was for the ID.ID case?

lvalueVariable
: (
// Check in semantic phase that the first id is not global/local/session/default.
identifier dotIdentifier?
| {this.isServerVersionGe80017()}? lValueIdentifier dotIdentifier?
)
| DEFAULT_SYMBOL dotIdentifier
;

and didnt adding a DEFAULT_SYMBOL dotIdentifier case just contradicted it?

[kaby76]

I suggest raising an Issue in the mysql-shell-plugins repo. The grammar here is scraped and refactored to target agnostic format from https://github.com/mysql/mysql-shell-plugins/tree/8928ada7d9e37a4075291880738983752b315fee/gui/frontend/src/parsing/mysql.