Teradata SQL Grammar #4330
Teradata SQL Grammar #4330
Conversation
| @@ -0,0 +1,31 @@ | |||
| # Teradata SQL Grammar | |||
|
|
|||
| An [ANTLR4](https://www.antlr.org/) grammar for Teradata SQL. Based on a grammar of Teradata Database version 17.10. | |||
There was a problem hiding this comment.
Please include a permalink to the actual grammar for version 17.10. Is it a bison grammar? Is there a lex file?
There was a problem hiding this comment.
Sorry, I don't understand what do you mean. Should I include links to formal grammar? I have none, and haven't found it on RDBMS vendor's website. I've written this grammar from the official publicly available documentation.
What information should I include in my case in readme? Links to documentation for version 17.10?
There was a problem hiding this comment.
Any of these?
- https://docs.teradata.com/r/Enterprise_IntelliFlex_VMware/SQL-Fundamentals/Basic-SQL-Syntax/SQL-Statement-Structure. The problem with this link is that the version "floats". There's no fixed version of the SQL syntax here. People can sneak in a change without anyone detecting the change.
- Is there anything under any of these Github repos to a pdf? https://github.com/Teradata
- Is there a PDF that contains text stating "version 17.10 of the SQL"?
- Does wayback.archive.org have a link to the doc fixed in time?
There was a problem hiding this comment.
I've added links to 17.10 docs. Those docs URLs end with /July-2021 subpath.
| numeric_data_type | ||
| : BYTEINT | ||
| | SMALLINT | ||
| | (INTEGER|INT) |
There was a problem hiding this comment.
Parentheses are used in places that don't need them. Why? E.g., (INTEGER | INT) instead of INTEGER | INT.
sql/teradata/TeradataSQLDDLParser.g4
Outdated
| cursor_declaration* | ||
| condition_handler* | ||
| procedure_stat* | ||
| END (label_name)? |
There was a problem hiding this comment.
Ditto on (label_name)?. Why this instead of label_name??
There was a problem hiding this comment.
Fixed. It looks like i've copied (label_name ':')? from the beginning of the same compound_stat rule and just removed ':'.
| (ON logging_item (',' logging_item)* )? | ||
| ; | ||
|
|
||
| operation |
There was a problem hiding this comment.
Please sort and uniq this list. You have duplicates, which is causing ambiguity to be detected in examples/ddl/logging/begin_logging.sql. For example, two INSERT, two SELECT, two UPDATE.
There was a problem hiding this comment.
Fixed. Thank you for review, I've copied list for operation rule from docs and haven't noticed that it had duplicates in sub-lists of common logging operations and row-level security logging operations.
|
(On the build, the MacOS runners in GitHub are terrible--hanging in the testing for Java. I'm not sure when they will finish, but it's stalled in parsing for the Java target. But, that works in about 9s on my machine just fine. My PR #4327 removes the parsing tests for MacOS in order to minimize hangs like this from happening.) |
DmitryMikhailovich
force-pushed
the
teradata-sql-grammar
branch
from
a88f95b to
8ffc041
Compare
Grammar of Teradata SQL dialect based on version 17.10 of Teradata Vantage.
Based on my estimates, this grammar is 90% complete. We used it to parse codebase of tens of thousands of objects for enterprise data lineage.
I've started to add examples only at a later stage of the development, so
examplesdir lacks many essential statements, like CRUD.Q: Why "Teradata SQL" is in naming, not just "Teradata"?
A: I've made it deliberately, because this grammar covers only SQL, without BTEQ scripting language (I think BTEQ scripts should have separate grammar).