trino-dev-cheat-sheet.md

Source: https://gist.github.com/findepi/04c96f0f60dcc95329f569bb0c44a0cd .

clean build

This should build the project. In case of problems, retry without the -T flag. This is not the fastest way to build the project though. See below.

./mvnw -T1C clean install -DskipTests 

use mvnd for development

When developing, mvnd is recommended, for faster feedback loops. You can install it with

# the v2 is not stable yet, hence @1 version pinning
brew install mvndaemon/homebrew-mvnd/mvnd@1

If you don't want to use mvnd, just replace it with ./mvnw -T1C.

quick clean build

mvnd clean install -nsu -DskipTests -Dmaven.javadoc.skip=true -Dmaven.source.skip=true -Dair.check.skip-all=true -pl '!:trino-docs,!:trino-proxy,!:trino-verifier,!:trino-benchto-benchmarks'

for even quicker incremental build, omit the clean

run static code analysis

# checkstyle only
mvnd checkstyle:check@checkstyle

# all verifications, including Error Prone compiler
# clean is needed to force recompilation with Error Prone compiler. 
# Re-run the second command with `-r` param to resume incremental build e.g. after fixing an error.
mvnd clean &&
mvnd test-compile install -DskipTests -Dmaven.javadoc.skip=true -Dmaven.source.skip=true -Dair.check.skip-all=true -Dskip.npm -Dskip.yarn -nsu -P errorprone-compiler -pl '!:trino-server,!:trino-docs'

code cleanups

mvnd license:format sortpom:sort
# I wish imports and formatting could be fixed from a CLI too

quick build antlr grammar alone

mvnd antlr4:antlr4 -pl :trino-grammar -nsu

# or all the grammars in the project
mvnd antlr4:antlr4 -nsu -pl "$(find . -name \*.g4 | sed -e 's@/src/.*@@' -e 's@.*/@:@' -e 's@$@,@' | sort -u)"

run Trino in a container, quickly

docker rm -f trino; docker run --rm -it --name trino -p 8080:8080 trinodb/trino

kill and remove all docker containers

docker ps -a | tail -n+2 | grep . && docker ps -aq | xargs docker rm -f ; docker volume prune -f

docs

docs/build

extract exception stacktrace from query JSON

# Just top-level exception
jq -r ' .failureInfo | [ .type + ": " + .message, (.stack[] | "    at " + .) ][] ' < query.json

# Full stacktrace
python -c '
import sys
import json
def print_exc(exc, enclosing_stacktrace=(), caption="", prefix=""):
    print("%s%s%s: %s" % (prefix, caption, exc["type"], exc["message"]))
    if "errorCode" in exc:
        print("%s\t\tcode: %s, name: %s" % (prefix, exc["errorCode"]["code"],
            exc["errorCode"]["name"]))
    stack_trace = exc["stack"]
    our_stack = unique_stacktrace(stack_trace, enclosing_stacktrace)
    for l in our_stack:
        print("%s\tat %s" % (prefix, l))
    if len(our_stack) < len(stack_trace):
        print("%s\t..." % (prefix, ))
    for s in exc["suppressed"]:
        print_exc(s, stack_trace, "Suppressed: ", prefix + "\t")
    if "cause" in exc:
        print_exc(exc["cause"], stack_trace, "Caused by: ", prefix)
def unique_stacktrace(stack, enclosing):
    stack = list(stack)
    enclosing = list(enclosing)
    while stack and enclosing and stack[-1] == enclosing[-1]:
        stack.pop()
        enclosing.pop()
    return stack
qj = json.load(sys.stdin)
print_exc(qj["failureInfo"])
' \
 < query.json

restore Product Test Launcher (ptl) removed by a very important person

# this is required for commands show below
git restore --source=$(git log -1 --format=tformat:%H -- bin/ptl)^ -- bin/ptl

product test env just as product tests do

bin/ptl env up --environment singlenode

postgresql

bin/ptl env up --environment singlenode-postgresql --without-trino
# docker exec -itu postgres ptl-postgresql psql -U test -d test

sqlserver

bin/ptl env up --environment singlenode-sqlserver --without-trino
# docker exec -it ptl-sqlserver /opt/mssql-tools/bin/sqlcmd -S localhost -U sa -P SQLServerPass1

mysql

bin/ptl env up --environment singlenode-mysql --without-trino
# docker exec -it ptl-mysql mysql -u test -ptest -D test

git bisect with plain test

git bisect run bash -xeuc '
  ./mvnw clean install -am -pl :trino-tests -T2C -DskipTests \
    -Dmaven.javadoc.skip=true -Dmaven.source.skip=true -Dair.check.skip-all=true || exit 125
  ./mvnw test -pl :trino-tests -Dair.check.skip-all=true -Dtest=TestLocalQueriesA#test
'

git bisect with product tests

git bisect run bash -xeuc '
  ./mvnw clean install -am -pl \!:trino-docs -DskipTests \
    -Dair.check.skip-all=true -Dmaven.javadoc.skip=true -Dmaven.source.skip=true -T2C || exit 125
  bin/ptl test run --environment singlenode -- -g tpcds -x quarantine -t q98
'

run product test quick

# TODO update this
#   java -jar presto-product-tests/target/presto-product-tests-*-executable.jar -c presto-product-tests/src/main/resources/sql-tests -g tpch -t q17

bring up product test env

# restore Product Test Launcher (ptl) removed by a very important person
git restore --source=$(git log -1 --format=tformat:%H -- bin/ptl)^ -- bin/ptl

# get a list of envs
bin/ptl env list

# start an env of your choice, exposing JVM debug port for Trino
bin/ptl env up --environment singlenode-hdp3 --debug

# run Trino CLI against the env
client/trino-cli/target/trino-cli-*-executable.jar --debug --server localhost:8080

# run Hive beeline 
docker exec -itu hive ptl-hadoop-master bash -l
echo "press Ctrl+R for bash history search and type 'beeline'"

# run Spark SQL (when dealing with an environment containing Spark image)
docker exec -it ptl-spark spark-sql
# ... and starting in some specific catalog
docker exec -it ptl-spark spark-sql -- --conf spark.sql.defaultCatalog=iceberg_test

check all commits compile

git rebase master -x './mvnw clean package -pl \!:trino-docs -T2C -DskipTests -Dmaven.javadoc.skip=true -Dmaven.source.skip=true'

sort commits for GitHub

git rebase master -x 'git commit --amend -C HEAD --date="$(date -R)" && sleep 1.05'

rebase commits within PR

This makes GitHub's diff for force-push useful.

git rebase -i master --keep-base

run benchmark from CLI

./mvnw exec:exec -pl :trino-main -Dexec.classpathScope=test -Dexec.executable="java" \
    -Dexec.args="-cp %classpath io.trino.operator.output.BenchmarkPartitionedOutputOperator"

network grepping

sudo ngrep -s0 -W byline -d lo0

AI

Cloud Code on the Web

Cloud Code on the Web's secure environment does not work out of the box.

Network access

Custom + add allowed domains

• repo.maven.apache.org

Environment variables

MAVEN_ARGS=-s /root/.maven.settings.xml -T1C
JAVA_HOME=/usr/lib/jvm/jdk-25.0.2-oracle-x64

Setup script

#!/bin/bash
set -eu

# Install system dependencies
export DEBIAN_FRONTEND=noninteractive
apt update
apt install -y curl wget

# Install JDK 25
wget https://download.oracle.com/java/25/latest/jdk-25_linux-x64_bin.deb
dpkg -i jdk-25_linux-x64_bin.deb
export JAVA_HOME=/usr/lib/jvm/jdk-25.0.2-oracle-x64

# Install some system cert into JDK to allow forward proxy (MITM)
keytool -importcert -trustcacerts \
  -keystore $JAVA_HOME/lib/security/cacerts \
  -storepass changeit -noprompt \
  -alias swp-ca-production \
  -file /usr/local/share/ca-certificates/swp-ca-production.crt

# Extract forward proxy configuration
extract() {
  local prop="$1"
  # Match -Dprop=value where value is a non-space token
  echo "$JAVA_TOOL_OPTIONS" | grep -oP "(?<=-D${prop}=)\S+" || echo ""
}
HTTP_HOST=$(extract http.proxyHost)
HTTP_PORT=$(extract http.proxyPort)
HTTP_USER=$(extract http.proxyUser)
HTTP_PASS=$(extract http.proxyPassword)
HTTPS_HOST=$(extract https.proxyHost)
HTTPS_PORT=$(extract https.proxyPort)
HTTPS_USER=$(extract https.proxyUser)
HTTPS_PASS=$(extract https.proxyPassword)
NON_PROXY=$(extract http.nonProxyHosts)

# Configure Maven to use proxy
cat > "$HOME/.maven.settings.xml" <<EOF
<?xml version="1.0" encoding="UTF-8"?>
<settings xmlns="http://maven.apache.org/SETTINGS/1.2.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://maven.apache.org/SETTINGS/1.2.0 https://maven.apache.org/xsd/settings-1.2.0.xsd">
  <proxies>
    <proxy>
      <id>http-proxy</id>
      <active>true</active>
      <protocol>http</protocol>
      <host>${HTTP_HOST}</host>
      <port>${HTTP_PORT}</port>
      <username>${HTTP_USER}</username>
      <password>${HTTP_PASS}</password>
      <nonProxyHosts>${NON_PROXY}</nonProxyHosts>
    </proxy>
    <proxy>
      <id>https-proxy</id>
      <active>true</active>
      <protocol>https</protocol>
      <host>${HTTPS_HOST}</host>
      <port>${HTTPS_PORT}</port>
      <username>${HTTPS_USER}</username>
      <password>${HTTPS_PASS}</password>
      <nonProxyHosts>${NON_PROXY}</nonProxyHosts>
    </proxy>
  </proxies>
</settings>
EOF

# Configure docker pull to use proxy
mkdir -p "$HOME/.docker"
cat > "$HOME/.docker/config.json" << EOF
{
  "proxies": {
    "default": {
      "httpProxy": "$HTTP_HOST",
      "httpsProxy": "$HTTP_HOST",
      "noProxy": "localhost,127.0.0.1"
    }
  }
}
EOF

# Docker is installed but not started by default
HTTP_PROXY="$HTTP_PROXY" HTTPS_PROXY="$HTTP_PROXY" dockerd &

echo Setup script all OK

I've been using this gist for years.

Google "findepi mvnw quick" then find this gist to copy the first command. Any way we could hack mvnw to create a standard build option @findepi @hashhar? I just want an --quick option that just does a default opinionated run like this :)

I just want an --quick option that just does a default opinionated run like this :)

That is what Maven profiles are for. We can add a profile in root pom (or you can add one in your personal settings.xml) with some options. For example I have in my settings.xml a profile that forces Maven Central to be tried first.

Would it make sense to have the profile in the default pom be set up for various builds like this? The default would be whatever it does today, and then you could have a quick build profile that runs a similar command that Piotr has above?

More bulletproof version for extracting stacktrace from query json.
Allows for either full query json or just part with faiulre. Also allows some garbage around in case you feed it with log line which just contains json payload you care about:

#!/usr/bin/env python3

import sys
import json
import re

def print_exc(exc, enclosing_stacktrace=(), caption="", prefix=""):
    if ("failureInfo" in exc):
        # if no failureInfo key assume we should treat whole jscon as failure description
        exc = exc["failureInfo"]

    print("%s%s%s: %s" % (prefix, caption, exc["type"], exc["message"]))
    if "errorCode" in exc:
        print("%s\t\tcode: %s, name: %s" % (prefix, exc["errorCode"]["code"],
            exc["errorCode"]["name"]))
    stack_trace = exc["stack"]
    our_stack = unique_stacktrace(stack_trace, enclosing_stacktrace)
    for l in our_stack:
        print("%s\tat %s" % (prefix, l))
    if len(our_stack) < len(stack_trace):
        print("%s\t..." % (prefix, ))
    for s in exc["suppressed"]:
        print_exc(s, stack_trace, "Suppressed: ", prefix + "\t")
    if "cause" in exc:
        print_exc(exc["cause"], stack_trace, "Caused by: ", prefix)
def unique_stacktrace(stack, enclosing):
    stack = list(stack)
    enclosing = list(enclosing)
    while stack and enclosing and stack[-1] == enclosing[-1]:
        stack.pop()
        enclosing.pop()
    return stack


tl_json = re.search("(\\{.*\\})", sys.stdin.read()).group(1)
qj = json.loads(tl_json)
print_exc(qj)

@losipiuk that's awesome!