/go-zetasqlfmt

Primary LanguageGo

go-zetasqlfmt

go-zetasqlfmt is a tool for formatting ZetaSQL within Go code.

It formats the SQL of the cloud.google.com/go/spanner.Statement structure using go-zetasql.

Installation

go install github.com/nametake/go-zetasqlfmt/cmd/zetasqlfmt@latest

If you don't want to see warning logs, you can suppress them by specifying CXXFLAGS as follows:

CGO_CXXFLAGS="$(go env CGO_CXXFLAGS) -Wno-deprecated" go install github.com/nametake/go-zetasqlfmt/cmd/zetasqlfmt@latest

If you use Docker, you can use the following commands without installing zetasqlfmt:

docker run --rm -v ".:/app" ghcr.io/nametake/go-zetasqlfmt:main ./...

Usage

You can format ZetaSQL by running the command within a project containing cloud.google.com/go/spanner.Statement.

Example code to be formatted:

package testdata

import (
	"cloud.google.com/go/spanner"
)

func Foo() *spanner.Statement {
	return &spanner.Statement{
		SQL:    "SELECT * FROM TABLE;",
		Params: map[string]interface{}{},
	}
}

Command execution:

zetasqlfmt ./...

Formatted code:

package testdata

import (
	"cloud.google.com/go/spanner"
)

func Foo() *spanner.Statement {
	return &spanner.Statement{
		SQL: `
SELECT
  *
FROM
  TABLE;
`, Params: map[string]interface{}{},
	}
}

Options

-nosemicolon
      no semicolon

Formatting Specifications

  • go-zetasql is used for SQL formatting.
  • If backticks ` are present in the formatted SQL, it removes line breaks and compresses multiple spaces into a single space.
  • If backticks ` are not present but line breaks are, it wraps the SQL with ` and line breaks.
  • If the SQL is formatted using fmt.Sprintf, it replaces it with dummy values before formatting.

In the case of dynamically using ORDER BY with cloud.google.com/go/spanner, a method using fmt.Sprintf has been suggest.

Therefore, in go-zetasqlfmt, in cases like the one below, it replaces format verbs with dummy values before formatting:

func Foo() *spanner.Statement {
	return &spanner.Statement{
		SQL:    fmt.Sprintf("SELECT * FROM TABLE ORDER BY %s;", "CreatedAt"),
		Params: map[string]interface{}{},
	}
}

The SQL is transformed as follows during formatting:

SELECT * FROM TABLE ORDER BY _DUMMY_STRING_;

The conversion table is as follows:

Verbs Dummy String
%s _DUMMY_STRING_
%v _DUMMY_VALUE_
%d -999

The replaced dummy values are then converted back to the original format verbs after formatting. Therefore, if the original SQL contains strings from the conversion table, unintended behavior may occur.

Links