You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
is_printable_as_block_string checks each character in a Python loop #312
print_schema() and print_ast() call is_printable_as_block_string() once for each description. The function walks the string one character at a time in a Python for loop. It keeps four boolean flags for each character. For long descriptions, this loop is a large part of the time that print_schema() uses.
Measured with Python 3.12 on one isolated CPU core (best of 7 runs):
Situation
Now
With str methods
Change
One-line description (30 chars)
1405 ns
184 ns
87% less time
Description with 12 lines
19.2 µs
2.7 µs
86% less time
print_schema() of the GitHub schema fixture
25.5 ms
16.3 ms
36% less time
String that starts with a new line
106 ns
427 ns
0.3 µs slower
The last row is the only slower case. The old loop stops at the first character there. Such strings are rare in real schemas.
Proposed fix: keep the same checks, but use str methods instead of the character loop. A str.isprintable() pre-check and a regex find control characters. Then str.split("\n") gives the lines for the leading, trailing and indent checks. The result stays the same for all inputs, and tests/language/test_block_string.py passes.
This is a Python-specific change only. It keeps the four rules of isPrintableAsBlockString in GraphQL.js as four short checks with the same comments, so a later sync with upstream stays easy.
Environment: graphql-core main (1d505c1), CPython 3.12.13, Linux x86_64.
Activity
perf(block-string): keep original loop with single-line shortcut
Cito commentedon Nov 15, 2026
Fixed by #313.