schemathesis_datagen_pipeline: state the Target 1 module interface

edgebench/tasks/schemathesis_datagen_pipeline/NOTES.md

@@ -24,7 +24,7 @@ Sibling of `schemathesis_config_modernization` (same task authors, same recipe).
   (hatchling 1.30.1). The editable install comes out byte-identical to v1's. Base: `bases/python`.
 - `judge/structured_result.py`, `judge/write_result.py`, `judge/run.sh`: v1's eval program and
   parser, shared with the sibling task.
-- `instruction.md`: v1's prompt, verbatim.
+- `instruction.md`: v1's prompt, verbatim in 1.0.0; 1.0.1 adds the Target 1 interface.
 
 ## Deviations from v1
 
@@ -52,6 +52,16 @@ range only from 0.3235 to 0.6718; the submissions below cover that range.
 
 Every row is valid, matches exactly with zero spread, and has identical per-phase summaries.
 
+## Changes
+
+### 1.0.1
+
+`TODO.md` does not give the module path, names or format keys that hidden Phase 1 imports, or the
+literal Range forms it checks, so a module that follows the requirements with other names scores 0
+of 15 on Phase 1 (15.38 of 100 points). `instruction.md` now states them. The judge
+and work images are unchanged (same tags), so grading is unchanged. `TODO.md` stays as in v1
+because the work image pins v1's commit.
+
 ## Known issues
 
 Kept as in v1, so scores stay comparable. Fixing any of them changes scores and bumps `version`.

edgebench/tasks/schemathesis_datagen_pipeline/instruction.md

@@ -6,6 +6,20 @@ You are an expert software engineer implementing a roadmap task in the current r
 
 Read `/app/TODO.md` completely before editing code. It contains the full project overview, goals, target requirements, and completion criteria.
 
+## Target 1 Module Interface
+
+`/app/TODO.md` does not name the following; they are part of Target 1. The new module is `schemathesis.specs.openapi.headers`, and it exposes:
+
+- `KNOWN_HEADER_FORMATS`: dict from lowercase header name to format key. `if-match` and `if-none-match` map to `_if_match_header`, `if-modified-since` and `if-unmodified-since` to `_http_date_header`, and `range` to `_range_header`.
+- `if_match_values()`, `http_date_values()`, `range_values()`, `range_slightly_invalid_values()`: each takes no arguments and returns a Hypothesis strategy of strings.
+- `get_header_format_strategies(mode)`: takes a `GenerationMode` and returns a dict from those three format keys to strategies.
+
+`if_match_values()` lists allow spaces after a comma, but none before a comma or around the whole list. `range_values()` serializes as lowercase `bytes=` followed by comma-separated range specs with no spaces between them.
+
+`range_slightly_invalid_values()` generates four forms: inverted bounds (`bytes=5-2`), a negative first position (`bytes=-1-N`), a wrong unit starting with `invalid=` (`invalid=0-100`), and the exact empty range set `bytes=`.
+
+In negative mode, the `_range_header` strategy mixes a valid Range value, the exact value `bytes=`, and a random value that starts with neither `bytes=` nor `invalid=`.
+
 ## Workflow
 
 1. Read the task document before making code changes.
@@ -17,4 +31,4 @@ Read `/app/TODO.md` completely before editing code. It contains the full project
 
 - Treat `/app/TODO.md` as the authoritative task specification.
 - Do not modify hidden tests.
-- Do not rely on external task text beyond the task document.
+- Do not rely on external task text beyond the task document and the interface above.

edgebench/tasks/schemathesis_datagen_pipeline/task.toml

@@ -4,7 +4,7 @@ name = "Schemathesis Datagen Pipeline"
 category = "Systems & Software Engineering"
 language = "Python"
 description = "Implement eight feature targets in the Schemathesis Python API testing framework, including structured HTTP header generation strategies, coverage-phase hooks, discriminator-aware validation and data generation, and schema-driven code generation fixes."
-version = "1.0.0"
+version = "1.0.1"
 status = "migrated"
 
 [environment]