vertica-sql-go is a native Go adapter for the Vertica (http://www.vertica.com) database.

Please check out release notes to learn about the latest improvements.

vertica-sql-go has been tested with Vertica 24.4.0 and Go 1.18/1.19/1.20/1.21/1.22.

Source code for vertica-sql-go can be found at:

https://github.com/vertica/vertica-sql-go

Alternatively you can use the 'go get' variant to install the package into your local Go environment.

go get github.com/vertica/vertica-sql-go

As this library is written to Go's SQL standard database/sql, usage is compliant with its methods and behavioral expectations.

First ensure that you have the library checked out in your standard Go hierarchy and import it.

import (
    "context"
    "database/sql"
    "github.com/vertica/vertica-sql-go"
)

The vertica-sql-go driver supports multiple log levels, as defined in the following table

and they can be set programmatically by calling the logger global level itself

logger.SetLogLevel(logger.DEBUG)

or by setting the environment variable VERTICA_SQL_GO_LOG_LEVEL to one of the integer values in the table above. This must be done before the process using the driver has started as the global log level will be read from here on start-up.

Example:

export VERTICA_SQL_GO_LOG_LEVEL=3

By default, log messages are sent to stdout, but the vertica-sql-go driver can also output to a file in cases where stdout is not available. Simply set the environment variable VERTICA_SQL_GO_LOG_FILE to your desired output location.

Example:

export VERTICA_SQL_GO_LOG_FILE=/var/log/vertica-sql-go.log

connDB, err := sql.Open("vertica", myDBConnectString)

where myDBConnectString is of the form:

vertica://(user):(password)@(host):(port)/(database)[?arg1=value&...&argN=valueN]

All parameters must be escaped (to URL-encoded format). If the host is a literal IPv6 address it must be enclosed in square brackets.

Currently supported query arguments are:

To ping the server and validate a connection (as the connection isn't necessarily created at that moment), simply call the PingContext() method.

ctx := context.Background()

err = connDB.PingContext(ctx)

If there is an error in connection, the error result will be non-nil and contain a description of whatever problem occurred.

Custom TLS config(s) can be registered for TLS / SSL encrypted connection to the server. Here is an example of registering and using a tls.Config:

import vertigo "github.com/vertica/vertica-sql-go"

// Register tls.Config
rootCertPool := x509.NewCertPool()
pem, err := ioutil.ReadFile("/certs/ca.crt")
if err != nil {
    LOG.Warningln("ERROR: failed reading cert file", err)
}
if ok := rootCertPool.AppendCertsFromPEM(pem); !ok {
    LOG.Warningln("ERROR: Failed to append PEM")
}
tlsConfig := &tls.Config{RootCAs: rootCertPool, ServerName: host}
vertigo.RegisterTLSConfig("myCustomName", tlsConfig)

// Connect using tls.Config
var rawQuery = url.Values{}
rawQuery.Add("tlsmode", "myCustomName")
var query = url.URL{
    Scheme:   "vertica",
    User:     url.UserPassword(user, password),
    Host:     fmt.Sprintf("%s:%d", host, port),
    Path:     databaseName,
    RawQuery: rawQuery.Encode(),
}
sql.Open("vertica", query.String())

Performing a simple query is merely a matter of using that connection to create a query and iterate its results. Here is an example of a query that should always work.

rows, err := connDB.QueryContext(ctx, "SELECT * FROM v_monitor.cpu_usage LIMIT 5")

defer rows.Close()

IMPORTANT : Just as with connections, you should always Close() the results cursor once you are done with it. It's often easier to just defer the closure, for convenience.

Prerequisites: Only SQL literals (i.e. query values) should be bound as arguments: they shouldn’t be used to merge table or field names to the query (vertica-sql-go will try quoting the table name as a string value, generating invalid SQL as it is actually a SQL Identifier). If you need to generate dynamically SQL queries (for instance choosing dynamically a table name) you have to construct the full query yourself.

rows, err := connDB.QueryContext(ctx, "SELECT name FROM MyTable WHERE id=?", 21)

Behind the scenes, this will be handled in one of two ways, based on whether or not you requested client interpolation in the connection string.

With client interpolation enabled, the client library will create a new query string with the arguments already in place, and submit it as a simple query.

With client interpolation disabled (default), the client library will use the full server-side parse(), describe(), bind(), execute() cycle.

rows, err := connDB.QueryContext(ctx, "SELECT name FROM MyTable WHERE id=@id and something=@example", sql.Named("id", 21), sql.Named("example", "hello"))

Named arguments are emulated by the driver. They will be converted to positional arguments by the driver and the named arguments given later will be slotted into the required positions. This still allows server side prepared statements as @id and @example above will be replaced by ? before being sent. If you use named arguments, all the arguments must be named. Do not mix positional and named together. All named arguments are normalized to upper case which means @param, @PaRaM, and @PARAM are treated as equivalent.

As outlined in the GoLang specs, reading the results of a query is done via a loop, bounded by a .next() iterator.

for rows.Next() {
    var nodeName string
    var startTime string
    var endTime string
    var avgCPU float64

    rows.Scan(&nodeName, &startTime, &endTime, &avgCPU)

    // Use these values for something here.
}

If you need to examine the names of the columns, simply access the Columns() operator of the rows object.

columnNames, _ := rows.Columns()

for _, columnName := range columnNames {
        // use the column name here.
}

By default, the query results are cached in memory allowing for rapid iteration of result row content. This generally works well, but in the case of exceptionally large result sets, you could run out of memory.

If such a query needs to be performed, it is recommended that you tell the driver that you wish to cache that data in a temporary file, so its results can be "paged in" as you iterate the results. The data is stored in a process-read-only file in the OS's temp directory.

To enable result paging, simply create a VerticaContext and use it to perform your query.

vCtx := NewVerticaContext(context.Background())

// Only keep 50000 rows in memory at once.
vCtx.SetInMemoryResultRowLimit(50000)

rows, _ := connDB.QueryContext(
    vCtx,
    "SELECT a, b, c, d, e FROM result_cache_test ORDER BY a")

defer rows.Close()

// Use rows result as normal.

If you want to disable paging on the same context all together, you can simply set the row limit to 0 (the default).

This is very similar to a simple query, but has a slightly different result type. A simple execute() might look like this:

res, err = connDB.ExecContext(ctx, "DROP TABLE IF EXISTS MyTable")

In this instance, res will contain information (such as 'rows affected') about the result of this execution.

This, again, looks very similar to the query-with-arguments use case and is subject to the same effects of client-side interpolation.

res, err := connDB.ExecContext(
        ctx,
        "INSERT INTO MyTable VALUES (?)", 21)

IMPORTANT : Vertica does not support executing a command string containing multiple statements using server-side prepared statements.

If you wish to reuse queries or executions, you can prepare them once and supply arguments only.

// Prepare the query.
stmt, err := connDB.PrepareContext(ctx, "SELECT id FROM MyTable WHERE name=?")

// Execute it with this argument.
rows, err = stmt.Query("Joe Perry")

NOTE : Please note that this method is subject to modification by the 'interpolate' setting. If the client side interpolation is requested, the statement will simply be stored on the client and interpolated with arguments each time it's used. If not using client side interpolation (default), the statement will be parsed and described on the server as expected.

The vertica-sql-go driver supports basic transactions as defined by the GoLang standard.

// Define the options for this transaction state
opts := &sql.TxOptions{
    Isolation: sql.LevelDefault,
    ReadOnly:  false,
}

// Begin the transaction.
tx, err := connDB.BeginTx(ctx, opts)

// You can either commit it.
err = tx.Commit()

// Or roll it back.
err = tx.Rollback()

The following transaction isolation levels are supported:

• sql.LevelReadUncommitted ^†
• sql.LevelReadCommitted
• sql.LevelSerializable
• sql.LevelRepeatableRead ^†
• sql.LevelDefault

The following transaction isolation levels are unsupported:

• sql.LevelSnapshot
• sql.LevelLinearizable

† Although Vertica supports the grammars for these transaction isolation levels, they are internally promoted to stronger isolation levels.

vertica-sql-go supports copying from stdin. This allows you to write a command-line tool that accepts stdin as an input and passes it to Vertica for processing. An example:

_, err = connDB.ExecContext(ctx, "COPY stdin_data FROM STDIN DELIMITER ','")

This will process input from stdin until an EOF is reached.

In your code, you may also supply a different io.Reader object (such as *File) from which to supply your data. Simply create a new VerticaContext, set the copy input stream, and provide this context to the execute call. An example:

fp, err := os.OpenFile("./resources/csv/sample_data.csv", os.O_RDONLY, 0600)
...
vCtx := NewVerticaContext(ctx)
vCtx.SetCopyInputStream(fp)

_, err = connDB.ExecContext(vCtx, "COPY stdin_data FROM STDIN DELIMITER ','")

If you provide a VerticaContext but don't set a copy input stream, the driver will fall back to os.stdin.

By following the above instructions, you should be able to successfully create a connection to your Vertica instance and perform the operations you require. A complete example program is listed below:

package main

import (
    "context"
    "database/sql"
    "os"

    _ "github.com/vertica/vertica-sql-go"
    "github.com/vertica/vertica-sql-go/logger"
)

func main() {
    // Have our logger output INFO and above.
    logger.SetLogLevel(logger.INFO)

    var testLogger = logger.New("samplecode")

    ctx := context.Background()

    // Create a connection to our database. Connection is lazy and won't
    // happen until it's used.
    connDB, err := sql.Open("vertica", "vertica://dbadmin:@localhost:5433/db1?connection_load_balance=1")

    if err != nil {
        testLogger.Fatal(err.Error())
        os.Exit(1)
    }

    defer connDB.Close()

    // Ping the database connnection to force it to attempt to connect.
    if err = connDB.PingContext(ctx); err != nil {
        testLogger.Fatal(err.Error())
        os.Exit(1)
    }

    // Query a standard metric table in Vertica.
    rows, err := connDB.QueryContext(ctx, "SELECT node_name, start_time, end_time, average_cpu_usage_percent FROM v_monitor.cpu_usage LIMIT 5")

    if err != nil {
        testLogger.Fatal(err.Error())
        os.Exit(1)
    }

    defer rows.Close()

    // Iterate over the results and print them out.
    for rows.Next() {
        var nodeName string
        var startTime string
        var endTime string
        var avgCPU float64

        if err = rows.Scan(&nodeName, &startTime, &endTime, &avgCPU); err != nil {
            testLogger.Fatal(err.Error())
            os.Exit(1)
        }

        testLogger.Info("%s\t%s\t%s\t%f", nodeName, startTime, endTime, avgCPU)
    }

    testLogger.Info("Test complete")

    os.Exit(0)
}

Apache 2.0 License, please see LICENSE for details.

Have a bug or an idea? Please see CONTRIBUTING.md for details.

You can run a benchmark and profile it with a command like: go test -bench '^BenchmarkRowsWithLimit$' -benchmem -memprofile memprofile.out -cpuprofile profile.out -run=none

and then explore it with go tool pprof. The -run part excludes the tests for brevity.

• @grzm (Github)
• @watercraft (Github)
• @fbernier (Github)
• @mlh758 (Github) for the awesome work filling in and enhancing the driver in many important ways.
• Tom Wall (Vertica) for the infinite patience and deep knowledge.
• The creators and contributors of the vertica-python library, and members of the Vertica team, for their help in understanding the wire protocol.