Files
openidea/crates/ide-db/src/clickhouse.rs
Nguyen Thinh 75c4ca0be8 feat(database): search and pagination on result query (#1432)
Adds a search bar to query result grids (match case / words / regex, previous/next occurrence, Filter rows) and a pager for console results (first/previous/next/last, page size). Later pages re-run the read-only query with a skip, which every driver reads past without decoding, so only one page crosses IPC; Last counts rows with the statement wrapped so a trailing line comment can't swallow the closing parenthesis.

Closes #1431

Co-authored-by: kidp2h <[email protected]>
2026-11-05 12:54:15 +08:00

890 lines
39 KiB
Rust

use crate::conn::{QueryResult, ResultColumn, Setup};
use crate::source::{DataSource, DbSslMode, DbTxMode};
use ide_core::{CancellationToken, IdeError, IdeResult, error::ErrorKind};
use parking_lot::RwLock;
use reqwest::{Client, RequestBuilder, StatusCode, Url};
use std::sync::Arc;
use std::time::{Duration, Instant};
use tokio::sync::Mutex;
#[derive(Clone)]
pub struct ClickhousePool(Arc<State>);
struct State {
client: RwLock<Option<Client>>,
endpoint: Url,
user: String,
password: String,
database: String,
settings: Vec<(String, String)>,
read_only: bool,
session_id: String,
session: Mutex<bool>,
closed: CancellationToken,
}
/// The response format: a names line, a types line, then one JSON array per row, so rows past
/// the grid's limit are skipped without being parsed.
const FORMAT: &str = "JSONCompactEachRowWithNamesAndTypes";
pub(crate) fn transaction_error() -> IdeError {
IdeError::invalid("ClickHouse HTTP does not support SQL transactions; use auto-commit and run writes directly")
}
impl ClickhousePool {
pub(crate) async fn connect(source: &DataSource, password: Option<&str>, setup: Setup) -> IdeResult<Self> {
if source.options.transaction == DbTxMode::Manual {
return Err(transaction_error());
}
for (name, _) in &setup.params {
if reserved_setting(name) {
return Err(IdeError::invalid(format!("ClickHouse HTTP manages the {name} setting")));
}
}
let secure = matches!(source.ssl, DbSslMode::Require | DbSslMode::VerifyCa | DbSslMode::VerifyFull);
let tls_files = [&source.ssl_ca_file, &source.ssl_cert_file, &source.ssl_key_file];
if !secure && tls_files.iter().any(|p| !p.trim().is_empty()) {
return Err(IdeError::invalid(
"ClickHouse TLS certificate files require an SSL mode of Require, Verify CA, or Verify Full",
));
}
let host = source.host.trim();
if host.is_empty() || host.contains(['/', '@', '?', '#']) {
return Err(IdeError::invalid("ClickHouse host must contain only a hostname or IP address"));
}
let host = if host.contains(':') && !host.starts_with('[') { format!("[{host}]") } else { host.to_owned() };
let scheme = if secure { "https" } else { "http" };
let port = source.port.unwrap_or(if secure { 8443 } else { 8123 });
let endpoint = Url::parse(&format!("{scheme}://{host}:{port}/"))
.map_err(|_| IdeError::invalid("invalid ClickHouse HTTP host or port"))?;
let timeout = setup.connect_timeout.unwrap_or(Duration::from_secs(30));
let _ = rustls::crypto::ring::default_provider().install_default();
let mut builder = Client::builder()
.connect_timeout(timeout)
.pool_max_idle_per_host(4)
.redirect(reqwest::redirect::Policy::none())
.user_agent(concat!("OpenIDEA/", env!("CARGO_PKG_VERSION")));
let ca = source.ssl_ca_file.trim();
if !ca.is_empty() {
let pem = read_pem(ca, "CA certificate")?;
let certificates = reqwest::Certificate::from_pem_bundle(&pem)
.map_err(|_| IdeError::invalid("invalid ClickHouse CA certificate PEM"))?;
if certificates.is_empty() {
return Err(IdeError::invalid("ClickHouse CA certificate PEM contains no certificates"));
}
builder = builder.tls_certs_only(certificates);
}
if source.ssl == DbSslMode::VerifyCa {
if ca.is_empty() {
return Err(IdeError::invalid("ClickHouse Verify CA requires a CA certificate file"));
}
builder = builder.tls_danger_accept_invalid_hostnames(true);
}
let cert = source.ssl_cert_file.trim();
let key = source.ssl_key_file.trim();
if cert.is_empty() != key.is_empty() {
return Err(IdeError::invalid("ClickHouse client TLS requires both certificate and key files"));
}
if !cert.is_empty() {
let mut pem = read_pem(cert, "client certificate")?;
pem.push(b'\n');
pem.extend(read_pem(key, "client key")?);
let identity = reqwest::Identity::from_pem(&pem)
.map_err(|_| IdeError::invalid("invalid ClickHouse client certificate or key PEM"))?;
builder = builder.identity(identity);
}
let proxy = ide_core::proxy_settings::proxy();
let client = ide_http::client::with_proxy(builder, &proxy).build().map_err(http_error)?;
let database = if source.options.default_schema.trim().is_empty() {
source.database.trim()
} else {
source.options.default_schema.trim()
};
let pool = Self(Arc::new(State {
client: RwLock::new(Some(client)),
endpoint,
user: if source.user.is_empty() { "default".into() } else { source.user.clone() },
password: password.unwrap_or_default().to_owned(),
database: if database.is_empty() { "default".into() } else { database.into() },
settings: setup.params,
read_only: source.read_only,
session_id: uuid::Uuid::new_v4().to_string(),
session: Mutex::new(false),
closed: CancellationToken::new(),
}));
let cancel = CancellationToken::new();
let initialize = async {
pool.execute("SELECT 1", 1, &cancel).await?;
for sql in setup.statements {
pool.execute(&sql, 0, &cancel).await?;
}
Ok(pool.clone())
};
tokio::pin!(initialize);
tokio::select! {
result = &mut initialize => {
if result.is_err() {
pool.close();
}
result
},
() = tokio::time::sleep(timeout) => {
cancel.cancel();
let _ = initialize.await;
pool.close();
Err(IdeError::new(ErrorKind::TimedOut, "timed out connecting to ClickHouse HTTP"))
}
}
}
pub fn close(&self) {
self.0.closed.cancel();
self.0.client.write().take();
}
pub fn is_closed(&self) -> bool {
self.0.closed.is_cancelled()
}
fn client(&self) -> IdeResult<Client> {
self.0.client.read().clone().ok_or_else(|| IdeError::new(ErrorKind::Io, "ClickHouse connection is closed"))
}
/// `max_rows`: the rows the server may send (the skipped ones included).
fn request(&self, client: &Client, query_id: &str, max_rows: usize, initialized: bool) -> RequestBuilder {
let mut request = client
.post(self.0.endpoint.clone())
.basic_auth(&self.0.user, Some(&self.0.password))
.header(reqwest::header::CONTENT_TYPE, "text/plain; charset=utf-8")
.query(&self.0.settings)
.query(&[
("session_id", self.0.session_id.clone()),
("session_timeout", "3600".into()),
("query_id", query_id.to_owned()),
("default_format", FORMAT.into()),
("output_format_json_quote_denormals", "1".into()),
("max_result_rows", max_rows.saturating_add(1).to_string()),
("result_overflow_mode", "break".into()),
]);
if initialized {
request = request.query(&[("session_check", "1")]);
} else {
request = request.query(&[("database", &self.0.database)]);
}
if self.0.read_only {
request = request.query(&[("readonly", "1")]);
}
request
}
pub async fn execute(&self, sql: &str, max_rows: usize, cancel: &CancellationToken) -> IdeResult<QueryResult> {
self.execute_from(sql, 0, max_rows, cancel).await
}
/// [`execute`](Self::execute), reading past the first `skip` rows without parsing or keeping them.
pub async fn execute_from(
&self,
sql: &str,
skip: usize,
max_rows: usize,
cancel: &CancellationToken,
) -> IdeResult<QueryResult> {
if transaction_statement(sql) {
return Err(transaction_error());
}
let mut session = tokio::select! {
biased;
() = cancel.cancelled() => return Err(IdeError::cancelled()),
() = self.0.closed.cancelled() => return Err(IdeError::new(ErrorKind::Io, "ClickHouse connection is closed")),
session = self.0.session.lock() => session,
};
let client = self.client()?;
let query_id = uuid::Uuid::new_v4().to_string();
let request = self.request(&client, &query_id, skip.saturating_add(max_rows), *session).body(sql.to_owned());
let started = Instant::now();
let fetch = async {
let mut response = request.send().await.map_err(http_error)?;
let execution_ms = started.elapsed().as_secs_f64() * 1000.0;
let status = response.status();
let header = |name: &str| response.headers().get(name).and_then(|v| v.to_str().ok()).map(str::to_owned);
let exception_code = header("X-ClickHouse-Exception-Code");
let format = header("X-ClickHouse-Format");
let affected = header("X-ClickHouse-Summary")
.and_then(|v| serde_json::from_str::<serde_json::Value>(&v).ok())
.and_then(|v| {
v.get("written_rows")
.and_then(|n| n.as_str().and_then(|n| n.parse::<u64>().ok()).or_else(|| n.as_u64()))
});
if !status.is_success() || exception_code.as_deref().is_some_and(|code| code != "0") {
let body = response.bytes().await.map_err(http_error)?;
return Err(self.server_error(status, exception_code.as_deref(), &body));
}
let (columns, rows, has_more) = if format.as_deref().is_some_and(|f| f != FORMAT) {
let body = response.bytes().await.map_err(http_error)?;
if !body.iter().all(u8::is_ascii_whitespace) {
return Err(IdeError::invalid(
"ClickHouse result grids need the default output format; remove the query's FORMAT clause",
));
}
(Vec::new(), Vec::new(), false)
} else {
let mut reader = RowReader::new(skip, max_rows);
while let Some(chunk) = response.chunk().await.map_err(http_error)? {
reader.push(&chunk)?;
}
reader.finish().map_err(|error| self.redact_error(error))?
};
let affected = if columns.is_empty() { affected.map(|n| n as f64) } else { None };
Ok(QueryResult {
statement: sql.to_owned(),
columns,
rows,
affected,
has_more,
elapsed_ms: started.elapsed().as_secs_f64() * 1000.0,
execution_ms,
})
};
let result = tokio::select! {
biased;
() = cancel.cancelled() => None,
() = self.0.closed.cancelled() => None,
result = fetch => Some(result),
};
match result {
Some(result) => {
if result.is_ok() {
*session = true;
}
result
}
None => {
let killed = self.kill(&client, &query_id).await;
match killed {
Ok(()) => Err(IdeError::cancelled()),
Err(error) => Err(IdeError::new(
ErrorKind::Cancelled,
format!("cancelled locally; ClickHouse server cancellation failed: {}", error.message),
)),
}
}
}
}
async fn kill(&self, client: &Client, query_id: &str) -> IdeResult<()> {
let response = client
.post(self.0.endpoint.clone())
.basic_auth(&self.0.user, Some(&self.0.password))
.timeout(Duration::from_secs(3))
.body(format!("KILL QUERY WHERE query_id = '{query_id}' SYNC"))
.send()
.await
.map_err(http_error)?;
let status = response.status();
let code = response
.headers()
.get("X-ClickHouse-Exception-Code")
.and_then(|value| value.to_str().ok())
.map(str::to_owned);
let body = response.bytes().await.map_err(http_error)?;
if status.is_success() && !code.as_deref().is_some_and(|code| code != "0") && !body.starts_with(b"Code:") {
Ok(())
} else {
Err(self.server_error(status, code.as_deref(), &body))
}
}
fn redact_error(&self, mut error: IdeError) -> IdeError {
if !self.0.password.is_empty() {
error.message = error.message.replace(&self.0.password, "[redacted]");
}
error
}
fn server_error(&self, status: StatusCode, code: Option<&str>, body: &[u8]) -> IdeError {
let message = String::from_utf8_lossy(body);
let message: String = exception_text(&message).chars().take(16_384).collect();
let auth = matches!(status, StatusCode::UNAUTHORIZED | StatusCode::FORBIDDEN)
|| matches!(code, Some("193" | "194" | "195" | "516" | "497"));
let detail = if message.is_empty() { status.canonical_reason().unwrap_or("request failed") } else { &message };
let detail = if detail.contains("SESSION_NOT_FOUND") || detail.contains("Session not found") {
"ClickHouse HTTP session expired; reconnect to restore session settings"
} else {
detail
};
self.redact_error(IdeError::new(
if auth { ErrorKind::PermissionDenied } else { ErrorKind::InvalidInput },
format!("ClickHouse HTTP {}: {detail}", status.as_u16()),
))
}
}
type DecodedResult = (Vec<ResultColumn>, Vec<Vec<Option<String>>>, bool);
/// Decodes a streamed [`FORMAT`] body. Reads past the first `skip` rows unparsed (a later
/// page), then keeps `max_rows` rows; later rows only set `has_more` and are read past
/// unparsed, so a grid limit bounds memory however large the block
/// `result_overflow_mode=break` lets the server send. A line that is not a JSON array
/// starts the exception ClickHouse writes when a query fails mid-stream.
struct RowReader {
skip: usize,
max_rows: usize,
pending: Vec<u8>,
names: Option<Vec<String>>,
types: Option<Vec<String>>,
rows: Vec<Vec<Option<String>>>,
more: bool,
exception: Option<String>,
}
impl RowReader {
fn new(skip: usize, max_rows: usize) -> Self {
Self {
skip,
max_rows,
pending: Vec::new(),
names: None,
types: None,
rows: Vec::new(),
more: false,
exception: None,
}
}
fn push(&mut self, chunk: &[u8]) -> IdeResult<()> {
self.pending.extend_from_slice(chunk);
let buffer = std::mem::take(&mut self.pending);
let mut start = 0;
while let Some(end) = buffer[start..].iter().position(|&b| b == b'\n') {
self.line(&buffer[start..start + end])?;
start += end + 1;
}
self.pending = buffer[start..].to_vec();
Ok(())
}
fn line(&mut self, line: &[u8]) -> IdeResult<()> {
if let Some(exception) = &mut self.exception {
if exception.len() < 16_384 {
exception.push_str(&String::from_utf8_lossy(line));
exception.push('\n');
}
return Ok(());
}
let text = line.trim_ascii();
if text.is_empty() {
return Ok(());
}
if text[0] != b'[' {
self.exception = Some(format!("{}\n", String::from_utf8_lossy(line)));
return Ok(());
}
if self.names.is_some() && self.types.is_some() {
if self.skip > 0 {
self.skip -= 1;
return Ok(());
}
if self.rows.len() >= self.max_rows {
self.more = true;
return Ok(());
}
}
let Ok(values) = serde_json::from_slice::<Vec<Box<serde_json::value::RawValue>>>(text) else {
self.exception = Some(format!("{}\n", String::from_utf8_lossy(line)));
return Ok(());
};
let protocol = |message: String| IdeError::new(ErrorKind::Protocol, message);
let header = |values: Vec<Box<serde_json::value::RawValue>>| {
values
.iter()
.map(|v| serde_json::from_str::<String>(v.get()))
.collect::<Result<Vec<_>, _>>()
.map_err(|e| protocol(format!("invalid ClickHouse result header: {e}")))
};
let Some(names) = &self.names else {
self.names = Some(header(values)?);
return Ok(());
};
if self.types.is_none() {
self.types = Some(header(values)?);
return Ok(());
}
if values.len() != names.len() {
return Err(protocol("ClickHouse result row width does not match its metadata".into()));
}
let row = values
.into_iter()
.map(|value| {
let text = value.get();
if text == "null" {
Ok(None)
} else if text.starts_with('"') {
serde_json::from_str::<String>(text).map(Some).map_err(|e| protocol(e.to_string()))
} else {
Ok(Some(text.to_owned()))
}
})
.collect::<IdeResult<_>>()?;
self.rows.push(row);
Ok(())
}
fn finish(mut self) -> IdeResult<DecodedResult> {
let rest = std::mem::take(&mut self.pending);
self.line(&rest)?;
if let Some(exception) = self.exception {
return Err(IdeError::invalid(format!("ClickHouse: {}", exception_text(&exception))));
}
let Some(names) = self.names else { return Ok((Vec::new(), Vec::new(), false)) };
let types = self.types.unwrap_or_default();
if types.len() != names.len() {
return Err(IdeError::new(ErrorKind::Protocol, "ClickHouse result types do not match its columns"));
}
let columns = names.into_iter().zip(types).map(|(name, type_name)| ResultColumn { name, type_name }).collect();
Ok((columns, self.rows, self.more))
}
}
/// The server's message in a failed response: rows already streamed and the
/// `__exception__` block markers ClickHouse wraps a mid-stream error in are dropped.
fn exception_text(body: &str) -> &str {
let Some(at) = body.find("Code:") else { return body.trim() };
let text = &body[at..];
match text.find("\n__exception__") {
// The line before the closing marker is `<length> <tag>`.
Some(end) => text[..end].trim_end().rsplit_once('\n').map_or(text[..end].trim(), |(message, _)| message.trim()),
None => text.trim(),
}
}
fn http_error(error: reqwest::Error) -> IdeError {
use std::error::Error;
let kind = if error.is_timeout() { ErrorKind::TimedOut } else { ErrorKind::Io };
let error = error.without_url();
let mut message = error.to_string();
let mut source = error.source();
while let Some(error) = source {
let detail = error.to_string();
if !message.contains(&detail) {
message.push_str(": ");
message.push_str(&detail);
}
source = error.source();
}
IdeError::new(kind, format!("ClickHouse HTTP: {message}"))
}
fn read_pem(path: &str, label: &str) -> IdeResult<Vec<u8>> {
std::fs::read(ide_ssh::expand(path))
.map_err(|error| IdeError::new(ErrorKind::Io, format!("cannot read ClickHouse {label}: {error}")))
}
fn reserved_setting(name: &str) -> bool {
let name = name.to_ascii_lowercase();
name.starts_with("output_format_json_")
|| matches!(
name.as_str(),
"user"
| "password"
| "query"
| "database"
| "session_id"
| "session_check"
| "session_timeout"
| "query_id"
| "default_format"
| "max_result_rows"
| "result_overflow_mode"
| "readonly"
)
}
fn transaction_statement(sql: &str) -> bool {
let mut sql = sql.trim_start();
loop {
if let Some(rest) = sql.strip_prefix("--").or_else(|| sql.strip_prefix('#')) {
sql = rest.split_once('\n').map_or("", |(_, rest)| rest).trim_start();
} else if let Some(rest) = sql.strip_prefix("/*") {
sql = rest.split_once("*/").map_or("", |(_, rest)| rest).trim_start();
} else {
break;
}
}
let mut words = sql.split(|c: char| !c.is_ascii_alphanumeric() && c != '_').filter(|s| !s.is_empty());
let first = words.next().unwrap_or_default();
["BEGIN", "COMMIT", "ROLLBACK", "SAVEPOINT", "RELEASE"].iter().any(|s| first.eq_ignore_ascii_case(s))
|| (first.eq_ignore_ascii_case("START") && words.next().is_some_and(|s| s.eq_ignore_ascii_case("TRANSACTION")))
}
#[cfg(test)]
mod tests {
use super::*;
use crate::conn::Pool;
use crate::source::{DbDriver, DbOptions, DbProperty};
use tokio::io::{AsyncReadExt, AsyncWriteExt};
const ONE: &str = "[\"1\"]\n[\"UInt8\"]\n[1]\n";
async fn read_request(stream: &mut tokio::net::TcpStream) -> String {
let mut bytes = Vec::new();
let mut buffer = [0; 4096];
loop {
let count = tokio::time::timeout(Duration::from_secs(5), stream.read(&mut buffer)).await.unwrap().unwrap();
assert!(count > 0, "request ended early");
bytes.extend_from_slice(&buffer[..count]);
if let Some(end) = bytes.windows(4).position(|s| s == b"\r\n\r\n") {
let header = String::from_utf8_lossy(&bytes[..end]);
let length: usize = header
.lines()
.find_map(|line| {
line.to_ascii_lowercase().strip_prefix("content-length:").map(|v| v.trim().parse().unwrap())
})
.unwrap_or(0);
if bytes.len() >= end + 4 + length {
return String::from_utf8(bytes).unwrap();
}
}
}
}
async fn reply(stream: &mut tokio::net::TcpStream, status: &str, headers: &str, body: &str) {
let response =
format!("HTTP/1.1 {status}\r\nContent-Length: {}\r\nConnection: close\r\n{headers}\r\n{body}", body.len());
stream.write_all(response.as_bytes()).await.unwrap();
}
async fn server(
responses: Vec<(&'static str, &'static str, &'static str)>,
) -> (u16, tokio::task::JoinHandle<Vec<String>>) {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let port = listener.local_addr().unwrap().port();
let task = tokio::spawn(async move {
let mut requests = Vec::new();
for (status, headers, body) in responses {
let (mut stream, _) =
tokio::time::timeout(Duration::from_secs(5), listener.accept()).await.unwrap().unwrap();
requests.push(read_request(&mut stream).await);
reply(&mut stream, status, headers, body).await;
}
requests
});
(port, task)
}
fn source(port: u16) -> DataSource {
DataSource { driver: DbDriver::Clickhouse, host: "127.0.0.1".into(), port: Some(port), ..Default::default() }
}
fn decode(body: &[u8], max_rows: usize, chunk: usize) -> IdeResult<DecodedResult> {
decode_from(body, 0, max_rows, chunk)
}
fn decode_from(body: &[u8], skip: usize, max_rows: usize, chunk: usize) -> IdeResult<DecodedResult> {
let mut reader = RowReader::new(skip, max_rows);
for part in body.chunks(chunk) {
reader.push(part)?;
}
reader.finish()
}
#[test]
fn rows_preserve_precision_nulls_and_strings_across_chunks() {
let body = concat!(
r#"["n", "d", "v", "v", "v"]"#,
"\n",
r#"["UInt64", "Decimal(38, 9)", "Nullable(String)", "Nullable(String)", "Nullable(String)"]"#,
"\n",
r#"[18446744073709551615, 12345678901234567890.123456789, null, "null", "ᴺᵁᴸᴸ"]"#,
);
for chunk in [1, 7, body.len()] {
let (columns, rows, more) = decode(body.as_bytes(), 1, chunk).unwrap();
assert_eq!(columns.len(), 5);
assert_eq!(columns[1].type_name, "Decimal(38, 9)");
assert_eq!(
rows[0],
vec![
Some("18446744073709551615".into()),
Some("12345678901234567890.123456789".into()),
None,
Some("null".into()),
Some("ᴺᵁᴸᴸ".into()),
]
);
assert!(!more);
}
let (columns, rows, more) = decode(b"[\"x\"]\n[\"String\"]\n", 0, 64).unwrap();
assert_eq!(columns[0].name, "x");
assert!(rows.is_empty() && !more);
assert_eq!(decode(b"", 5, 64).unwrap().0.len(), 0);
assert!(decode(b"[\"x\"]\n[\"UInt8\"]\n[1, 2]\n", 1, 64).is_err());
}
#[test]
fn a_later_page_skips_rows_without_keeping_them() {
let body = b"[\"n\"]\n[\"UInt8\"]\n[1]\n[2]\n[3]\n[4]\n[5]\n";
let page = |skip, max| {
let (_, rows, more) = decode_from(body, skip, max, 3).unwrap();
(rows.into_iter().map(|r| r[0].clone().unwrap()).collect::<Vec<_>>(), more)
};
assert_eq!(page(2, 2), (vec!["3".to_owned(), "4".to_owned()], true));
assert_eq!(page(4, 2), (vec!["5".to_owned()], false));
assert_eq!(page(9, 2), (Vec::<String>::new(), false));
}
#[test]
fn rows_past_the_limit_are_skipped_but_a_later_exception_still_fails() {
let mut body = String::from("[\"n\"]\n[\"UInt64\"]\n");
for n in 0..1000 {
body.push_str(&format!("[{n}]\n"));
}
let (_, rows, more) = decode(body.as_bytes(), 2, 4096).unwrap();
assert_eq!(rows, vec![vec![Some("0".into())], vec![Some("1".into())]]);
assert!(more);
body.push_str("\n__exception__\nabcdefgh\nCode: 395. DB::Exception: Value passed to 'throwIf' is non-zero\n");
body.push_str("62 abcdefgh\n__exception__\n");
let error = decode(body.as_bytes(), 2, 4096).unwrap_err();
assert_eq!(error.message, "ClickHouse: Code: 395. DB::Exception: Value passed to 'throwIf' is non-zero");
let multiline = "Code: 62. DB::Exception: Syntax error\nat line 2";
assert_eq!(exception_text(multiline), multiline);
}
#[tokio::test]
async fn http_requests_use_auth_settings_limits_and_a_shared_session() {
let (port, requests) = server(vec![
("200 OK", "X-ClickHouse-Format: JSONCompactEachRowWithNamesAndTypes\r\n", ONE),
("200 OK", "", ""),
(
"200 OK",
"X-ClickHouse-Format: JSONCompactEachRowWithNamesAndTypes\r\n",
"[\"n\"]\n[\"UInt64\"]\n[\"1\"]\n[\"2\"]\n[\"3\"]\n",
),
])
.await;
let source = DataSource {
user: "reader".into(),
read_only: true,
options: DbOptions { default_schema: "analytics".into(), time_zone: "UTC".into(), ..Default::default() },
properties: vec![DbProperty { name: "max_threads".into(), value: "2".into() }],
..source(port)
};
let pool = Pool::connect(&source, Some("mock-password"), None).await.unwrap();
pool.execute("USE other", 0, &CancellationToken::new()).await.unwrap();
let result = pool.execute("SELECT number FROM numbers(3)", 1, &CancellationToken::new()).await.unwrap();
assert_eq!(result.rows, vec![vec![Some("1".into())]]);
assert!(result.has_more && result.affected.is_none());
let requests = requests.await.unwrap();
let target = |request: &String| {
Url::parse(&format!(
"http://localhost{}",
request.lines().next().unwrap().split_whitespace().nth(1).unwrap()
))
.unwrap()
};
let urls: Vec<_> = requests.iter().map(target).collect();
let value = |url: &Url, key: &str| url.query_pairs().find(|(k, _)| k == key).map(|(_, v)| v.into_owned());
assert_eq!(value(&urls[0], "database").as_deref(), Some("analytics"));
assert_eq!(value(&urls[1], "database"), None);
assert_eq!(value(&urls[1], "session_check").as_deref(), Some("1"));
assert_eq!(value(&urls[0], "session_id"), value(&urls[2], "session_id"));
assert_ne!(value(&urls[0], "query_id"), value(&urls[2], "query_id"));
assert_eq!(value(&urls[2], "max_result_rows").as_deref(), Some("2"));
for (request, url) in requests.iter().zip(&urls) {
assert!(request.starts_with("POST /?"));
assert!(request.contains("Basic cmVhZGVyOm1vY2stcGFzc3dvcmQ="));
assert!(!url.as_str().contains("reader") && !url.as_str().contains("mock-password"));
assert_eq!(value(url, "readonly").as_deref(), Some("1"));
assert_eq!(value(url, "session_timezone").as_deref(), Some("UTC"));
assert_eq!(value(url, "max_threads").as_deref(), Some("2"));
}
assert!(requests[0].ends_with("SELECT 1"));
assert!(pool.transaction(&[]).await.is_err());
let cancel = CancellationToken::new();
cancel.cancel();
assert_eq!(pool.execute("SELECT 1", 1, &cancel).await.unwrap_err().kind, ErrorKind::Cancelled);
pool.close().await;
assert!(pool.is_closed());
assert!(pool.execute("SELECT 1", 1, &CancellationToken::new()).await.is_err());
}
#[tokio::test]
async fn sequential_queries_reuse_the_http_connection() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let port = listener.local_addr().unwrap().port();
let task = tokio::spawn(async move {
let (mut stream, _) = listener.accept().await.unwrap();
for _ in 0..2 {
read_request(&mut stream).await;
let response = format!("HTTP/1.1 200 OK\r\nContent-Length: {}\r\n\r\n{ONE}", ONE.len());
stream.write_all(response.as_bytes()).await.unwrap();
}
});
let pool = Pool::connect(&source(port), None, None).await.unwrap();
assert_eq!(
pool.execute("SELECT 1", 1, &CancellationToken::new()).await.unwrap().rows,
vec![vec![Some("1".into())]]
);
task.await.unwrap();
}
#[tokio::test]
async fn server_errors_keep_details_and_hide_passwords() {
let (port, requests) = server(vec![(
"403 Forbidden",
"X-ClickHouse-Exception-Code: 516\r\n",
"Code: 516. DB::Exception: bad mock-password",
)])
.await;
let error = match Pool::connect(&source(port), Some("mock-password"), None).await {
Ok(_) => panic!("expected login failure"),
Err(error) => error,
};
assert_eq!(error.kind, ErrorKind::PermissionDenied);
assert!(error.message.contains("516") && error.message.contains("[redacted]"));
assert!(!error.message.contains("mock-password"));
requests.await.unwrap();
let (port, requests) = server(vec![("200 OK", "", "Code: 241. DB::Exception: memory limit")]).await;
let error = match Pool::connect(&source(port), None, None).await {
Ok(_) => panic!("expected late query failure"),
Err(error) => error,
};
assert!(error.message.contains("memory limit"));
requests.await.unwrap();
let (port, requests) = server(vec![(
"500 Internal Server Error",
"X-ClickHouse-Exception-Code: 395\r\n",
"[\"1\"]\n[\"UInt8\"]\n[1]\n\n__exception__\ntag\nCode: 395. DB::Exception: thrown\n32 tag\n__exception__\n",
)])
.await;
let error = match Pool::connect(&source(port), None, None).await {
Ok(_) => panic!("expected mid-stream failure"),
Err(error) => error,
};
assert_eq!(error.message, "ClickHouse HTTP 500: Code: 395. DB::Exception: thrown");
requests.await.unwrap();
}
#[tokio::test]
async fn cancellation_sends_kill_outside_the_busy_session() {
let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
let port = listener.local_addr().unwrap().port();
let (started, received) = tokio::sync::oneshot::channel();
let server = tokio::spawn(async move {
let (mut health, _) = listener.accept().await.unwrap();
read_request(&mut health).await;
reply(&mut health, "200 OK", "", ONE).await;
drop(health);
let (mut query, _) = listener.accept().await.unwrap();
let request = read_request(&mut query).await;
started.send(()).unwrap();
let (mut kill, _) = tokio::time::timeout(Duration::from_secs(5), listener.accept()).await.unwrap().unwrap();
let kill_request = read_request(&mut kill).await;
reply(&mut kill, "200 OK", "", "").await;
(request, kill_request)
});
let pool = Pool::connect(&source(port), None, None).await.unwrap();
let cancel = CancellationToken::new();
let query = pool.execute("SELECT sleep(30)", 1, &cancel);
let cancelled = async {
received.await.unwrap();
cancel.cancel();
};
let (result, ()) = tokio::join!(query, cancelled);
assert_eq!(result.unwrap_err().kind, ErrorKind::Cancelled);
let (query, kill) = server.await.unwrap();
let target = query.lines().next().unwrap().split_whitespace().nth(1).unwrap();
let url = Url::parse(&format!("http://localhost{target}")).unwrap();
let query_id = url.query_pairs().find(|(k, _)| k == "query_id").unwrap().1.into_owned();
assert!(kill.ends_with(&format!("KILL QUERY WHERE query_id = '{query_id}' SYNC")));
assert!(!kill.contains("session_id="));
}
#[tokio::test]
async fn unsupported_settings_and_transactions_fail_before_connecting() {
for statement in ["BEGIN", "-- comment\n COMMIT", "/* comment */ ROLLBACK", "START TRANSACTION", "SAVEPOINT s"]
{
assert!(transaction_statement(statement));
}
assert!(!transaction_statement("SELECT 'BEGIN'"));
let sources = [
DataSource { options: DbOptions { transaction: DbTxMode::Manual, ..Default::default() }, ..source(1) },
DataSource {
properties: vec![DbProperty { name: "statement_cache_capacity".into(), value: "4".into() }],
..source(1)
},
DataSource { properties: vec![DbProperty { name: "password".into(), value: "x".into() }], ..source(1) },
DataSource { ssl_ca_file: "unused.pem".into(), ..source(1) },
DataSource { ssl: DbSslMode::VerifyCa, ..source(1) },
DataSource { ssl: DbSslMode::Require, ssl_cert_file: "unused.pem".into(), ..source(1) },
];
for source in sources {
let error = match Pool::connect(&source, None, None).await {
Ok(_) => panic!("unsupported settings were accepted"),
Err(error) => error,
};
assert_eq!(error.kind, ErrorKind::InvalidInput);
}
}
#[tokio::test]
#[ignore = "requires CLICKHOUSE_TEST_HOST, CLICKHOUSE_TEST_USER, and CLICKHOUSE_TEST_PASSWORD"]
async fn clickhouse_cloud_transport() {
let source = DataSource {
driver: DbDriver::Clickhouse,
host: std::env::var("CLICKHOUSE_TEST_HOST").expect("CLICKHOUSE_TEST_HOST"),
port: Some(std::env::var("CLICKHOUSE_TEST_PORT").ok().and_then(|v| v.parse().ok()).unwrap_or(8443)),
user: std::env::var("CLICKHOUSE_TEST_USER").expect("CLICKHOUSE_TEST_USER"),
ssl: DbSslMode::Require,
options: DbOptions {
time_zone: "UTC".into(),
startup_script: "SET max_threads = 2".into(),
..Default::default()
},
properties: vec![DbProperty { name: "max_block_size".into(), value: "50".into() }],
..Default::default()
};
let password = std::env::var("CLICKHOUSE_TEST_PASSWORD").expect("CLICKHOUSE_TEST_PASSWORD");
let pool = Pool::connect(&source, Some(&password), None).await.unwrap();
let cancel = CancellationToken::new();
let settings = pool
.execute("SELECT getSetting('max_threads'), getSetting('max_block_size'), timezone()", 1, &cancel)
.await
.unwrap();
assert_eq!(settings.rows, vec![vec![Some("2".into()), Some("50".into()), Some("UTC".into())]]);
let result = pool.execute("SELECT toUInt64('18446744073709551615') AS n, CAST(NULL AS Nullable(String)) AS missing, CAST('ᴺᵁᴸᴸ' AS Nullable(String)) AS text, 'a\\nb' AS lines", 1, &cancel).await.unwrap();
assert_eq!(
result.rows,
vec![vec![Some("18446744073709551615".into()), None, Some("ᴺᵁᴸᴸ".into()), Some("a\nb".into())]]
);
assert_eq!(result.columns[1].type_name, "Nullable(String)");
let limited = pool.execute("SELECT number FROM numbers(100)", 2, &cancel).await.unwrap();
assert_eq!(limited.rows.len(), 2);
assert!(limited.has_more);
let marker = format!("transport_cancel_{}", uuid::Uuid::new_v4().simple());
let sql = format!("SELECT sleepEachRow(0.05) FROM numbers(50) /* {marker} */");
let stop = CancellationToken::new();
let query = pool.execute(&sql, 1, &stop);
let cancel_query = async {
tokio::time::sleep(Duration::from_millis(500)).await;
stop.cancel();
};
let (cancelled, ()) = tokio::join!(query, cancel_query);
let cancelled = cancelled.unwrap_err();
assert_eq!(cancelled.kind, ErrorKind::Cancelled);
assert_eq!(cancelled.message, "cancelled");
let active = pool.execute(&format!("SELECT count() FROM system.processes WHERE position(query, '{marker}') > 0 AND query_id != currentQueryID()"), 1, &cancel).await.unwrap();
assert_eq!(active.rows, vec![vec![Some("0".into())]]);
let table = format!("transport_test_{}", uuid::Uuid::new_v4().simple());
pool.execute(&format!("CREATE TEMPORARY TABLE {table} (n UInt8)"), 0, &cancel).await.unwrap();
let continuity = async {
pool.execute(&format!("INSERT INTO {table} VALUES (7)"), 0, &cancel).await?;
pool.execute(&format!("SELECT n FROM {table}"), 1, &cancel).await
}
.await;
let cleanup = pool.execute(&format!("DROP TEMPORARY TABLE {table}"), 0, &cancel).await;
assert_eq!(continuity.unwrap().rows, vec![vec![Some("7".into())]]);
cleanup.unwrap();
pool.close().await;
assert!(pool.is_closed());
}
}