A HyperDX row click finds its row, a source may alias its columns, and CSV answers (T-496) - #359
Merged
Conversation
chasers
added this pull request to stack #354
September 20, 2026 02:16
…d CSV answers (T-496)
Four things HyperDX does that the edge did not take, each checked against
HyperDX's own code or the statement it builds:
- WITH (expr) AS alias. HyperDX sends one per alias of a source's select
list (ServiceName as service), so a filter can say "service" in a
statement whose SELECT does not define it, as its histogram's does not.
The engine's WITH takes only subqueries. Rewrite drops those items from
the list, drops the list when nothing else is in it, and writes the alias
as its expression wherever it is read: not where a SELECT defines it, not
after a dot, not before a parenthesis.
- A row click. HyperDX finds the clicked row by the values it was shown:
ts=parseDateTime64BestEffort('...', 9) AND attrs=JSONExtract('{...}',
'Map(String, String)') AND ... It found nothing, for two reasons that
were the edge's answers, not its SQL. A TIMESTAMP_NS answered six digits,
because a NaiveDateTime holds six; Query now reads such a column as
integer nanoseconds and Format writes all nine. A map answered its keys
in Elixir's order, and the engine compares maps entry by entry, as
ClickHouse does; Frame.to_rows/2 takes map_entries: true and the edge
answers a map in the order it was stored. JSONExtract(json, 'Type')
becomes a cast, for scalars, Map(K, V) and Array(T). MD5 answers bytes,
as ClickHouse's does, since HyperDX wraps it in lower(hex(...)) for a
long string. toJSONString and leftUTF8 are macros.
- CSV, CSVWithNames and CSVWithNamesAndTypes, which HyperDX asks for alert
sample rows.
Verified on a local node with two rows differing only in their last three
nanosecond digits: HyperDX's row-click statement answers the one it was
built from. The alias histogram, the MD5 predicate and CSV answer through
HyperDX's client.
docs/clickstack.md gains a section on the SQLite catalog: HyperDX sends its
metadata queries in parallel, and on one node with a SQLite catalog that
meets smolquery's existing "database is locked" issue, which eight
concurrent queries through the plain API reproduce with no ClickHouse edge
running. A Postgres catalog does not have it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
… may use an earlier one, and a name defined without AS is not a call WITH (expr) AS alias was a flat word replace over the statement. Three statements the review ran showed what that misses: - A subquery that defines the name for itself. With WITH (ServiceName) AS service SELECT service FROM (SELECT ServiceName AS service ...) ORDER BY service, the outer "service" is the subquery's column, and writing (ServiceName) there named a column that is not in scope. An alias a subquery defines with AS is now not substituted anywhere, and no name is substituted inside a subquery. - Chained aliases. (a + 1) AS b, (b * 2) AS c left c as (b * 2), with b unresolved. Each expression now has the earlier aliases written into it. - An alias defined without AS. SELECT ServiceName service became ServiceName (ServiceName), a call. A name is read only where a value can stand: after a keyword such as SELECT, WHERE, AND or BY, an operator, a comma or an opening parenthesis; not after another name, a closing parenthesis, a quoted name or a literal. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
chasers
force-pushed
the
t-496-with-expression-alias
branch
from
September 20, 2026 03:11
9026a64 to
4129dde
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
TL;DR: A HyperDX row click now finds its row, a source may alias its columns (
ServiceName as service), andCSVoutput answers.Tracker: T-496. Plan PL-66. Stacked on #358.
What changed
WITH (expr) AS aliasfor an aliased select listCSV(alert sample rows)CSV,CSVWithNames,CSVWithNamesAndTypesWhy the row click found nothing
TIMESTAMP_NSanswered 6 digits. Elixir'sNaiveDateTimeholds 6. HyperDX sent the value back, and it no longer matched the stored 9.How it works
Queryreads a nanosecond column as integer nanoseconds.Formatwrites all 9 digits, in every format.Frame.to_rows/2takesmap_entries: true. The edge answers a map in its stored key order.Rewrite:WITH (expr) AS aliasitems leave theWITHlist. The alias is written as its expression wherever it is read.Rewrite:JSONExtract(json, 'Type')becomes a cast, for scalars,Map(K, V)andArray(T).Rewrite:MD5(x)answers bytes, as ClickHouse's does. HyperDX wraps it inlower(hex(...))for a long string.toJSONString,leftUTF8.Watch out
TIMESTAMP_NScolumn now answers 9 fractional digits on this edge. It was 6.SELECTdefines it, after a., or before a(.WITH expr AS aliaswithout parentheses is still a parse error. HyperDX always writes them.database is lockedissue. Eight concurrent queries through the plain API reproduce it with no ClickHouse edge running. A Postgres catalog does not have it.How it was verified
MD5predicate,CSV.Checks
mix precommitmix cimix dialyzerReview fixes
WITHalias is read in its own scope: not inside a subquery, and not at all when a subquery defines the name for itself.(a + 1) AS b, (b * 2) AS c.SELECT ServiceName service(noAS) is a definition, not a call.🤖 Generated with Claude Code