Formatted SQL adds spaces inside the next string literal after a string ending in a backslash

Aperta
  • Qualcosa non va?
#6,500 2 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 3 giorni

Nessuno ha ancora preso questa issue.

  • #6507 di @ayushdoesdev — chiusa senza merge

Valutazione

Difficoltà
4/5
Tempo stimato
1-2 giorni
Idoneità per principianti
38/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
rust, sql
Ambito
backend, databases

Direzione di ricerca

Start in prqlc/prqlc/src/sql/mod.rs at sql::compile and trace how generated SQL is passed to sqlformat::format; inspect the sqlformat tokenizer behavior described in the issue. Compare the proposed options and add a regression test for the backslash-ending string example. Done means formatted output preserves the literal value; the issue leaves the interim approach and upstream report for maintainers to decide.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

bug

When a string literal in the query ends with a backslash, the formatted SQL (the default output) changes the value of the next string literal by adding spaces inside its quotes. The query then compares against, or selects, a different string than the one written. --no-format output is correct.

prql target:sql.postgres
from files
filter path == 'C:\\' || name == 'readme'

prqlc compile --no-format, which is correct:

SELECT * FROM files WHERE path = 'C:\' OR name = 'readme'

prqlc compile:

SELECT
  *
FROM
  files
WHERE
  path = 'C:\' OR name = ' readme '

name = ' readme ' matches no row that name = 'readme' would. The same happens with derive {a = 'ab\\', b = 'c'}, which emits ' c ' AS b, and with raw strings like r'x\'. Any string whose value ends in a backslash triggers it; the query doesn't have to be unusual.

Cause: sql::compile passes the SQL that sqlparser generates through sqlformat::format. The sqlformat 0.5.0 tokenizer (get_string_token in src/tokenizer.rs) always treats a backslash as an escape inside '...' and "...". So it reads 'C:\' as an unterminated string that runs on to the next '. The quotes then pair up wrongly, and the formatter adds spaces around tokens that are really inside string literals. In standard SQL, and in the Postgres, DuckDB, SQLite and generic dialects PRQL targets, a backslash is an ordinary character, so 'C:\' is a complete literal. FormatOptions::dialect (Generic / PostgreSql / SQLServer) doesn't change this, and 0.5.0 is the latest sqlformat release. I found no report of this in shssoichiro/sqlformat-rs.

Why this is an issue rather than a PR: the durable fix is upstream, in sqlformat's tokenizer, perhaps keyed on its dialect. Any interim change in prqlc is a trade-off a maintainer should pick:

  • Skip formatting, and return the unformatted SQL, when the generated SQL contains a backslash. This is simple and can't corrupt values, but those queries lose their layout.
  • Mask string literals before formatting and restore them afterwards. This keeps the layout, but adds a second tokenizer pass in prqlc.

Either change would come with a regression test like the example above. Filing the upstream report would also need a decision: the repo shows no agent-contribution policy, so per this repo's tend guidance it waits for a go-ahead here.

Lingua principale
Rust
Stelle
10.9k
Fork
261
Merge medio
3g 20h
PR unite (30g)
85

Preparare l'ambiente

Apri in Codespaces

Avvia il container di sviluppo del progetto nel browser, con il tuo account GitHub.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di PRQL/prql

Tutte le issue di PRQL/prql

Issue simili

Altre issue su Rust

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.