What SQLite column name can be/cannot be?

Question

Is there any rule for the SQLite\'s column name?

• Can it have characters like \'/\'?
• Can it be UTF-8?

Answer 1

The following answer is based on the SQLite source code, mostly relying on the file parse.y (input for the lemon parser).

TL;DR:

The allowed series of characters for column and table names in CREATE TABLE statements are

• '-escaped strings of any kind (even keywords)
• Identifiers, which means
  • ``` and "-escaped strings of any kind (even keywords)
  • a series of the MSB=1 8-bit ASCII characters or 7-bit ASCII characters with 1 in the following table that doesn't form a keyword:
• Keyword INDEXED because it's non-standard
• Keyword JOIN for reason that is unknown to me.

The allowed series of characters for result columns in a SELECT statement are

• Either a string or an identifier as described above
• All of the above if used as a column alias written after AS

Now to the exploration process itself

1. let's look at the syntax for CREATE TABLE columns

   // The name of a column or table can be any of the following:
   //
   %type nm {Token}
   nm(A) ::= id(X).         {A = X;}
   nm(A) ::= STRING(X).     {A = X;}
   nm(A) ::= JOIN_KW(X).    {A = X;}
   

2. digging deeper, we find out that

   // An IDENTIFIER can be a generic identifier, or one of several
   // keywords.  Any non-standard keyword can also be an identifier.
   //
   %type id {Token}
   id(A) ::= ID(X).         {A = X;}
   id(A) ::= INDEXED(X).    {A = X;}
   

   "Generic identifier" sounds unfamiliar. A quick look into tokenize.c however brings forth the definition

   /*
   ** The sqlite3KeywordCode function looks up an identifier to determine if
   ** it is a keyword.  If it is a keyword, the token code of that keyword is 
   ** returned.  If the input is not a keyword, TK_ID is returned.
   */
   
   /*
   ** If X is a character that can be used in an identifier then
   ** IdChar(X) will be true.  Otherwise it is false.
   **
   ** For ASCII, any character with the high-order bit set is
   ** allowed in an identifier.  For 7-bit characters, 
   ** sqlite3IsIdChar[X] must be 1.
   **
   ** Ticket #1066.  the SQL standard does not allow '$' in the
   ** middle of identfiers.  But many SQL implementations do. 
   ** SQLite will allow '$' in identifiers for compatibility.
   ** But the feature is undocumented.
   */
   

   For a full map of identifier characters, please consult the tokenize.c.

3. It is still unclear what are the available names for a result-column (i. e. the column name or alias assigned in the SELECT statement). parse.y is again helpful here.

   // An option "AS " phrase that can follow one of the expressions that
   // define the result set, or one of the tables in the FROM clause.
   //
   %type as {Token}
   as(X) ::= AS nm(Y).    {X = Y;}
   as(X) ::= ids(Y).      {X = Y;}
   as(X) ::= .            {X.n = 0;}