openapi-spec-validator-rs 0.2.0__tar.gz → 0.2.2__tar.gz

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/Cargo.lock

@@ -848,7 +848,7 @@ checksum = "384b8ab6d37215f3c5301a95a4accb5d64aa607f1fcb26a11b5303878451b4fe"
 
 [[package]]
 name = "openapi-spec-validator-rs"
-version = "0.2.0"
+version = "0.2.2"
 dependencies = [
  "base64",
  "fancy-regex",
@@ -864,7 +864,7 @@ dependencies = [
 
 [[package]]
 name = "openapi-spec-validator-rs-cli"
-version = "0.2.0"
+version = "0.2.2"
 dependencies = [
  "clap",
  "insta",
@@ -875,7 +875,7 @@ dependencies = [
 
 [[package]]
 name = "openapi-spec-validator-rs-py"
-version = "0.2.0"
+version = "0.2.2"
 dependencies = [
  "clap",
  "mimalloc",

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/Cargo.toml

@@ -3,7 +3,7 @@ resolver = "3"
 members = ["crates/*"]
 
 [workspace.package]
-version = "0.2.0"
+version = "0.2.2"
 edition = "2024"
 rust-version = "1.85"
 license = "Apache-2.0"
@@ -14,8 +14,8 @@ keywords = ["openapi", "swagger", "validation", "jsonschema"]
 categories = ["command-line-utilities", "development-tools", "web-programming"]
 
 [workspace.dependencies]
-openapi-spec-validator-rs = { path = "crates/openapi-spec-validator-rs", version = "0.2.0", default-features = false }
-openapi-spec-validator-rs-cli = { path = "crates/openapi-spec-validator-rs-cli", version = "0.2.0", default-features = false }
+openapi-spec-validator-rs = { path = "crates/openapi-spec-validator-rs", version = "0.2.2", default-features = false }
+openapi-spec-validator-rs-cli = { path = "crates/openapi-spec-validator-rs-cli", version = "0.2.2", default-features = false }
 
 anyhow = "1.0.104"
 base64 = "0.23.1"

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openapi-spec-validator-rs
-Version: 0.2.0
+Version: 0.2.2
 Classifier: Development Status :: 4 - Beta
 Classifier: Intended Audience :: Developers
 Classifier: Operating System :: OS Independent
@@ -28,7 +28,7 @@ Project-URL: Repository, https://github.com/Arjun-Aravind/openapi-spec-validator
 **A fast, drop-in Rust companion to [openapi-spec-validator](https://github.com/python-openapi/openapi-spec-validator).**
 
 It validates Swagger 2.0 and OpenAPI 3.0, 3.1 and 3.2 documents with the same rules as
-openapi-spec-validator, returns the same verdicts, and runs 40 to 145 times faster on large
+openapi-spec-validator, returns the same verdicts, and runs 37 to 143 times faster on large
 specs. It ships as a command-line tool, a Python package with the same API, and a pre-commit
 hook. It is an independent project, not affiliated with or endorsed by the maintainers of
 openapi-spec-validator.
@@ -39,10 +39,10 @@ with its default backend):
 
 | Spec | openapi-spec-validator-rs | openapi-spec-validator | Speed-up |
 |---|---:|---:|---:|
-| GitHub (OpenAPI 3.0, JSON, 13.5 MB) | **57 ms** | 6.41 s – 8.18 s | **113–144×** |
-| GitHub (OpenAPI 3.0, YAML, 10.3 MB) | **152 ms** | 6.45 s – 8.20 s | **42–54×** |
-| Stripe (OpenAPI 3.0, JSON, 8.3 MB) | **31 ms** | 1.80 s – 2.82 s | **59–92×** |
-| Kubernetes (Swagger 2.0, JSON, 4.3 MB) | **20 ms** | 762 ms – 2.13 s | **39–109×** |
+| GitHub (OpenAPI 3.0, JSON, 13.5 MB) | **57 ms** | 6.45 s – 8.21 s | **112–143×** |
+| GitHub (OpenAPI 3.0, YAML, 10.3 MB) | **150 ms** | 6.46 s – 8.23 s | **43–55×** |
+| Stripe (OpenAPI 3.0, JSON, 8.3 MB) | **31 ms** | 1.81 s – 2.85 s | **59–93×** |
+| Kubernetes (Swagger 2.0, JSON, 4.3 MB) | **20 ms** | 766 ms – 2.14 s | **37–104×** |
 
 The `pip`-installed command runs the same validator behind a Python entry point, which adds
 about 25 ms of interpreter start-up per run: 45 ms for Kubernetes and 80 ms for GitHub's
@@ -113,14 +113,20 @@ The package mirrors openapi-spec-validator's API:
 ```python
 from openapi_spec_validator_rs import validate, validate_path, validate_url
 from openapi_spec_validator_rs import OpenAPIV31SpecValidator, OpenAPIValidationError
+from openapi_spec_validator_rs.readers import read_from_filename
 
 validate(spec_dict)                     # raises the first error
 validate_path("openapi.yaml")           # reads the file like the command line
 validate_url("https://example.com/openapi.json")
 
-errors = list(OpenAPIV31SpecValidator(spec_dict).iter_errors())
+spec_dict, base_uri = read_from_filename("openapi.yaml")
+errors = list(OpenAPIV31SpecValidator(spec_dict, base_uri=base_uri).iter_errors())
 ```
 
+`spec_dict` can be anything `json.load` or `yaml.safe_load` returns, dates and all: a
+`datetime.date` from an unquoted `2020-01-01` is treated as jsonschema treats it (of no
+type, shown as `datetime.date(2020, 1, 1)` in messages).
+
 Errors are `OpenAPIValidationError` (or one of its subclasses: `DuplicateOperationIDError`,
 `ParameterDuplicateError`, `UnresolvableParameterError`, `ExtraParametersError`) with the
 same attributes and messages as upstream's. Moving over is usually a matter of changing the
@@ -165,7 +171,7 @@ repos:
         name: openapi-spec-validator-rs
         entry: openapi-spec-validator-rs
         language: python
-        additional_dependencies: [openapi-spec-validator-rs==0.2.0]
+        additional_dependencies: [openapi-spec-validator-rs==0.2.2]
         files: .*(openapi|swagger).*\.(json|yaml|yml)$
 ```
 
@@ -184,12 +190,13 @@ validates FastAPI's `app.openapi()` does), median of 5 runs:
 
 | Spec | openapi-spec-validator-rs | openapi-spec-validator, jsonschema-rs backend | openapi-spec-validator, default backend |
 |---|---:|---:|---:|
-| GitHub (13.5 MB) | **33 ms** | 4.67 s (143× slower) | 6.38 s (196× slower) |
-| Stripe (8.3 MB) | **16 ms** | 0.89 s (55× slower) | 1.91 s (119× slower) |
-| Kubernetes (4.3 MB) | **11 ms** | 0.33 s (29× slower) | 1.70 s (149× slower) |
+| GitHub (13.5 MB) | **30 ms** | 4.66 s (158× slower) | 6.40 s (217× slower) |
+| Stripe (8.3 MB) | **15 ms** | 0.89 s (58× slower) | 1.92 s (126× slower) |
+| Kubernetes (4.3 MB) | **11 ms** | 0.33 s (30× slower) | 1.70 s (154× slower) |
 
-It also needs about a quarter of the memory on JSON specs: 81 MiB instead of 290 MiB for
-GitHub's (98 MiB through the `pip`-installed command).
+On the same machine it also needs about a quarter of the memory on JSON specs: 84 MiB
+instead of 289 MiB for GitHub's (98 MiB through the `pip`-installed command); the ratio
+varies with the platform's allocator and Python build.
 
 Validating every spec in the APIs.guru directory that openapi-spec-validator accepts (3,646
 files, 535 MB of YAML) takes 1.6 s instead of 202–332 s.
@@ -203,8 +210,10 @@ The validator is a Rust library (`crates/openapi-spec-validator-rs`) with a CLI
 PyO3 bindings on top. It follows the reference implementation step by step:
 
 1. **Load**: JSON goes through serde_json when libyaml would read it the same way; anything
-   else goes through libyaml (the parser inside PyYAML) with PyYAML's YAML 1.1 rules, then
-   through the same JSON round trip.
+   else goes through libyaml (the parser inside PyYAML, as the [unsafe-libyaml] crate's
+   mechanical translation of its C code: archived upstream, which is also what keeps it
+   identical to the parser PyYAML bundles) with PyYAML's YAML 1.1 rules, then through
+   the same JSON round trip. Streams are read only as far as that parser would get.
 2. **Meta-schema**: the whole document is validated against the official schema for its
    version with the [`jsonschema`](https://crates.io/crates/jsonschema) crate.
 3. **Walk**: the document is walked like `keywords.py`, following `$ref`s (into other
@@ -232,4 +241,5 @@ and [NOTICE][notice].
 [compatibility]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/COMPATIBILITY.md
 [license]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/LICENSE
 [notice]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/NOTICE
+[unsafe-libyaml]: https://crates.io/crates/unsafe-libyaml
 

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/README.md

@@ -3,7 +3,7 @@
 **A fast, drop-in Rust companion to [openapi-spec-validator](https://github.com/python-openapi/openapi-spec-validator).**
 
 It validates Swagger 2.0 and OpenAPI 3.0, 3.1 and 3.2 documents with the same rules as
-openapi-spec-validator, returns the same verdicts, and runs 40 to 145 times faster on large
+openapi-spec-validator, returns the same verdicts, and runs 37 to 143 times faster on large
 specs. It ships as a command-line tool, a Python package with the same API, and a pre-commit
 hook. It is an independent project, not affiliated with or endorsed by the maintainers of
 openapi-spec-validator.
@@ -14,10 +14,10 @@ with its default backend):
 
 | Spec | openapi-spec-validator-rs | openapi-spec-validator | Speed-up |
 |---|---:|---:|---:|
-| GitHub (OpenAPI 3.0, JSON, 13.5 MB) | **57 ms** | 6.41 s – 8.18 s | **113–144×** |
-| GitHub (OpenAPI 3.0, YAML, 10.3 MB) | **152 ms** | 6.45 s – 8.20 s | **42–54×** |
-| Stripe (OpenAPI 3.0, JSON, 8.3 MB) | **31 ms** | 1.80 s – 2.82 s | **59–92×** |
-| Kubernetes (Swagger 2.0, JSON, 4.3 MB) | **20 ms** | 762 ms – 2.13 s | **39–109×** |
+| GitHub (OpenAPI 3.0, JSON, 13.5 MB) | **57 ms** | 6.45 s – 8.21 s | **112–143×** |
+| GitHub (OpenAPI 3.0, YAML, 10.3 MB) | **150 ms** | 6.46 s – 8.23 s | **43–55×** |
+| Stripe (OpenAPI 3.0, JSON, 8.3 MB) | **31 ms** | 1.81 s – 2.85 s | **59–93×** |
+| Kubernetes (Swagger 2.0, JSON, 4.3 MB) | **20 ms** | 766 ms – 2.14 s | **37–104×** |
 
 The `pip`-installed command runs the same validator behind a Python entry point, which adds
 about 25 ms of interpreter start-up per run: 45 ms for Kubernetes and 80 ms for GitHub's
@@ -88,14 +88,20 @@ The package mirrors openapi-spec-validator's API:
 ```python
 from openapi_spec_validator_rs import validate, validate_path, validate_url
 from openapi_spec_validator_rs import OpenAPIV31SpecValidator, OpenAPIValidationError
+from openapi_spec_validator_rs.readers import read_from_filename
 
 validate(spec_dict)                     # raises the first error
 validate_path("openapi.yaml")           # reads the file like the command line
 validate_url("https://example.com/openapi.json")
 
-errors = list(OpenAPIV31SpecValidator(spec_dict).iter_errors())
+spec_dict, base_uri = read_from_filename("openapi.yaml")
+errors = list(OpenAPIV31SpecValidator(spec_dict, base_uri=base_uri).iter_errors())
 ```
 
+`spec_dict` can be anything `json.load` or `yaml.safe_load` returns, dates and all: a
+`datetime.date` from an unquoted `2020-01-01` is treated as jsonschema treats it (of no
+type, shown as `datetime.date(2020, 1, 1)` in messages).
+
 Errors are `OpenAPIValidationError` (or one of its subclasses: `DuplicateOperationIDError`,
 `ParameterDuplicateError`, `UnresolvableParameterError`, `ExtraParametersError`) with the
 same attributes and messages as upstream's. Moving over is usually a matter of changing the
@@ -140,7 +146,7 @@ repos:
         name: openapi-spec-validator-rs
         entry: openapi-spec-validator-rs
         language: python
-        additional_dependencies: [openapi-spec-validator-rs==0.2.0]
+        additional_dependencies: [openapi-spec-validator-rs==0.2.2]
         files: .*(openapi|swagger).*\.(json|yaml|yml)$
 ```
 
@@ -159,12 +165,13 @@ validates FastAPI's `app.openapi()` does), median of 5 runs:
 
 | Spec | openapi-spec-validator-rs | openapi-spec-validator, jsonschema-rs backend | openapi-spec-validator, default backend |
 |---|---:|---:|---:|
-| GitHub (13.5 MB) | **33 ms** | 4.67 s (143× slower) | 6.38 s (196× slower) |
-| Stripe (8.3 MB) | **16 ms** | 0.89 s (55× slower) | 1.91 s (119× slower) |
-| Kubernetes (4.3 MB) | **11 ms** | 0.33 s (29× slower) | 1.70 s (149× slower) |
+| GitHub (13.5 MB) | **30 ms** | 4.66 s (158× slower) | 6.40 s (217× slower) |
+| Stripe (8.3 MB) | **15 ms** | 0.89 s (58× slower) | 1.92 s (126× slower) |
+| Kubernetes (4.3 MB) | **11 ms** | 0.33 s (30× slower) | 1.70 s (154× slower) |
 
-It also needs about a quarter of the memory on JSON specs: 81 MiB instead of 290 MiB for
-GitHub's (98 MiB through the `pip`-installed command).
+On the same machine it also needs about a quarter of the memory on JSON specs: 84 MiB
+instead of 289 MiB for GitHub's (98 MiB through the `pip`-installed command); the ratio
+varies with the platform's allocator and Python build.
 
 Validating every spec in the APIs.guru directory that openapi-spec-validator accepts (3,646
 files, 535 MB of YAML) takes 1.6 s instead of 202–332 s.
@@ -178,8 +185,10 @@ The validator is a Rust library (`crates/openapi-spec-validator-rs`) with a CLI
 PyO3 bindings on top. It follows the reference implementation step by step:
 
 1. **Load**: JSON goes through serde_json when libyaml would read it the same way; anything
-   else goes through libyaml (the parser inside PyYAML) with PyYAML's YAML 1.1 rules, then
-   through the same JSON round trip.
+   else goes through libyaml (the parser inside PyYAML, as the [unsafe-libyaml] crate's
+   mechanical translation of its C code: archived upstream, which is also what keeps it
+   identical to the parser PyYAML bundles) with PyYAML's YAML 1.1 rules, then through
+   the same JSON round trip. Streams are read only as far as that parser would get.
 2. **Meta-schema**: the whole document is validated against the official schema for its
    version with the [`jsonschema`](https://crates.io/crates/jsonschema) crate.
 3. **Walk**: the document is walked like `keywords.py`, following `$ref`s (into other
@@ -207,3 +216,4 @@ and [NOTICE][notice].
 [compatibility]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/COMPATIBILITY.md
 [license]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/LICENSE
 [notice]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/NOTICE
+[unsafe-libyaml]: https://crates.io/crates/unsafe-libyaml

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/crates/openapi-spec-validator-rs/README.md

@@ -3,7 +3,7 @@
 **A fast, drop-in Rust companion to [openapi-spec-validator](https://github.com/python-openapi/openapi-spec-validator).**
 
 It validates Swagger 2.0 and OpenAPI 3.0, 3.1 and 3.2 documents with the same rules as
-openapi-spec-validator, returns the same verdicts, and runs 40 to 145 times faster on large
+openapi-spec-validator, returns the same verdicts, and runs 37 to 143 times faster on large
 specs. It ships as a command-line tool, a Python package with the same API, and a pre-commit
 hook. It is an independent project, not affiliated with or endorsed by the maintainers of
 openapi-spec-validator.
@@ -14,10 +14,10 @@ with its default backend):
 
 | Spec | openapi-spec-validator-rs | openapi-spec-validator | Speed-up |
 |---|---:|---:|---:|
-| GitHub (OpenAPI 3.0, JSON, 13.5 MB) | **57 ms** | 6.41 s – 8.18 s | **113–144×** |
-| GitHub (OpenAPI 3.0, YAML, 10.3 MB) | **152 ms** | 6.45 s – 8.20 s | **42–54×** |
-| Stripe (OpenAPI 3.0, JSON, 8.3 MB) | **31 ms** | 1.80 s – 2.82 s | **59–92×** |
-| Kubernetes (Swagger 2.0, JSON, 4.3 MB) | **20 ms** | 762 ms – 2.13 s | **39–109×** |
+| GitHub (OpenAPI 3.0, JSON, 13.5 MB) | **57 ms** | 6.45 s – 8.21 s | **112–143×** |
+| GitHub (OpenAPI 3.0, YAML, 10.3 MB) | **150 ms** | 6.46 s – 8.23 s | **43–55×** |
+| Stripe (OpenAPI 3.0, JSON, 8.3 MB) | **31 ms** | 1.81 s – 2.85 s | **59–93×** |
+| Kubernetes (Swagger 2.0, JSON, 4.3 MB) | **20 ms** | 766 ms – 2.14 s | **37–104×** |
 
 The `pip`-installed command runs the same validator behind a Python entry point, which adds
 about 25 ms of interpreter start-up per run: 45 ms for Kubernetes and 80 ms for GitHub's
@@ -88,14 +88,20 @@ The package mirrors openapi-spec-validator's API:
 ```python
 from openapi_spec_validator_rs import validate, validate_path, validate_url
 from openapi_spec_validator_rs import OpenAPIV31SpecValidator, OpenAPIValidationError
+from openapi_spec_validator_rs.readers import read_from_filename
 
 validate(spec_dict)                     # raises the first error
 validate_path("openapi.yaml")           # reads the file like the command line
 validate_url("https://example.com/openapi.json")
 
-errors = list(OpenAPIV31SpecValidator(spec_dict).iter_errors())
+spec_dict, base_uri = read_from_filename("openapi.yaml")
+errors = list(OpenAPIV31SpecValidator(spec_dict, base_uri=base_uri).iter_errors())
 ```
 
+`spec_dict` can be anything `json.load` or `yaml.safe_load` returns, dates and all: a
+`datetime.date` from an unquoted `2020-01-01` is treated as jsonschema treats it (of no
+type, shown as `datetime.date(2020, 1, 1)` in messages).
+
 Errors are `OpenAPIValidationError` (or one of its subclasses: `DuplicateOperationIDError`,
 `ParameterDuplicateError`, `UnresolvableParameterError`, `ExtraParametersError`) with the
 same attributes and messages as upstream's. Moving over is usually a matter of changing the
@@ -140,7 +146,7 @@ repos:
         name: openapi-spec-validator-rs
         entry: openapi-spec-validator-rs
         language: python
-        additional_dependencies: [openapi-spec-validator-rs==0.2.0]
+        additional_dependencies: [openapi-spec-validator-rs==0.2.2]
         files: .*(openapi|swagger).*\.(json|yaml|yml)$
 ```
 
@@ -159,12 +165,13 @@ validates FastAPI's `app.openapi()` does), median of 5 runs:
 
 | Spec | openapi-spec-validator-rs | openapi-spec-validator, jsonschema-rs backend | openapi-spec-validator, default backend |
 |---|---:|---:|---:|
-| GitHub (13.5 MB) | **33 ms** | 4.67 s (143× slower) | 6.38 s (196× slower) |
-| Stripe (8.3 MB) | **16 ms** | 0.89 s (55× slower) | 1.91 s (119× slower) |
-| Kubernetes (4.3 MB) | **11 ms** | 0.33 s (29× slower) | 1.70 s (149× slower) |
+| GitHub (13.5 MB) | **30 ms** | 4.66 s (158× slower) | 6.40 s (217× slower) |
+| Stripe (8.3 MB) | **15 ms** | 0.89 s (58× slower) | 1.92 s (126× slower) |
+| Kubernetes (4.3 MB) | **11 ms** | 0.33 s (30× slower) | 1.70 s (154× slower) |
 
-It also needs about a quarter of the memory on JSON specs: 81 MiB instead of 290 MiB for
-GitHub's (98 MiB through the `pip`-installed command).
+On the same machine it also needs about a quarter of the memory on JSON specs: 84 MiB
+instead of 289 MiB for GitHub's (98 MiB through the `pip`-installed command); the ratio
+varies with the platform's allocator and Python build.
 
 Validating every spec in the APIs.guru directory that openapi-spec-validator accepts (3,646
 files, 535 MB of YAML) takes 1.6 s instead of 202–332 s.
@@ -178,8 +185,10 @@ The validator is a Rust library (`crates/openapi-spec-validator-rs`) with a CLI
 PyO3 bindings on top. It follows the reference implementation step by step:
 
 1. **Load**: JSON goes through serde_json when libyaml would read it the same way; anything
-   else goes through libyaml (the parser inside PyYAML) with PyYAML's YAML 1.1 rules, then
-   through the same JSON round trip.
+   else goes through libyaml (the parser inside PyYAML, as the [unsafe-libyaml] crate's
+   mechanical translation of its C code: archived upstream, which is also what keeps it
+   identical to the parser PyYAML bundles) with PyYAML's YAML 1.1 rules, then through
+   the same JSON round trip. Streams are read only as far as that parser would get.
 2. **Meta-schema**: the whole document is validated against the official schema for its
    version with the [`jsonschema`](https://crates.io/crates/jsonschema) crate.
 3. **Walk**: the document is walked like `keywords.py`, following `$ref`s (into other
@@ -207,3 +216,4 @@ and [NOTICE][notice].
 [compatibility]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/COMPATIBILITY.md
 [license]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/LICENSE
 [notice]: https://github.com/Arjun-Aravind/openapi-spec-validator-rs/blob/main/NOTICE
+[unsafe-libyaml]: https://crates.io/crates/unsafe-libyaml

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/crates/openapi-spec-validator-rs/src/document.rs

@@ -19,6 +19,7 @@ use serde_json::Value;
 use url::Url;
 
 use crate::loader;
+use crate::pyobj;
 use crate::pyrepr;
 use crate::pyurl;
 use crate::retrieve::Retrieve;
@@ -123,6 +124,9 @@ pub(crate) struct Documents<'a> {
     externals: Arena<External>,
     /// Document index by URI (without fragment): 0 is the root, `n` is `externals[n - 1]`.
     by_uri: RefCell<HashMap<String, usize>>,
+    /// The one-character strings that pointers into strings resolve to: Python indexes
+    /// a `str` like a sequence and yields a new `str` each time.
+    characters: Arena<Value>,
 }
 
 impl<'a> Documents<'a> {
@@ -138,6 +142,7 @@ impl<'a> Documents<'a> {
             retriever,
             externals: Arena::new(),
             by_uri: RefCell::new(by_uri),
+            characters: Arena::new(),
         }
     }
 
@@ -223,7 +228,7 @@ impl<'a> Documents<'a> {
                 let path = url
                     .to_file_path()
                     .map_err(|()| format!("not a file path: {uri}"))?;
-                let bytes = std::fs::read(&path).map_err(|error| error.to_string())?;
+                let bytes = crate::read_file(&path).map_err(|error| error.to_string())?;
                 // Upstream reads files through `urlopen`, which is the stream name
                 // PyYAML reports.
                 loader::load(&bytes, "<urllib response>").map_err(|error| error.message)
@@ -296,14 +301,78 @@ impl<'a> Documents<'a> {
                 ),
             ));
         }
-        let value = pointer(contents, fragment)?;
+        let value = self.pointer(contents, fragment)?;
         Ok(Node { value, doc })
     }
 
+    /// `Resource.pointer` from `referencing`: each segment indexes a mapping by key or a
+    /// sequence by position. A Python `str` is a sequence too, so a pointer can reach
+    /// into a string and come out with one of its characters.
+    fn pointer<'d>(&'d self, document: &'d Value, pointer: &str) -> Result<&'d Value, LookupError> {
+        let nowhere = || {
+            let mut message = format!(
+                "{} does not exist within {}",
+                pyrepr::repr_str(pointer),
+                pyrepr::repr(document)
+            );
+            if pointer == "/" {
+                message.push_str(
+                    ". The pointer '/' is a valid JSON Pointer but it points to an empty string \
+                     property ''. If you intended to point to the entire resource, you should use '#'.",
+                );
+            }
+            LookupError::referencing("PointerToNowhere", message)
+        };
+        let invalid_index = |segment: &str| {
+            LookupError::from(Crash(format!(
+                "invalid literal for int() with base 10: {}",
+                pyrepr::repr_str(segment)
+            )))
+        };
+        // `int(segment)`, then Python's negative indexing.
+        let position = |segment: &str, len: usize| -> Result<Option<usize>, LookupError> {
+            let index = python_int(segment).ok_or_else(|| invalid_index(segment))?;
+            let index = if index < 0 {
+                i128::try_from(len).unwrap_or(i128::MAX) + index
+            } else {
+                index
+            };
+            Ok(usize::try_from(index).ok().filter(|&i| i < len))
+        };
+        let decoded = percent_decode_str(&pointer[1..]).decode_utf8_lossy();
+        let mut contents = document;
+        for segment in decoded.split('/') {
+            contents = match contents {
+                Value::Array(items) => position(segment, items.len())?
+                    .map(|i| &items[i])
+                    .ok_or_else(nowhere)?,
+                Value::Object(map) if pyrepr::foreign(contents).is_none() => {
+                    let key = segment.replace("~1", "/").replace("~0", "~");
+                    map.get(&key).ok_or_else(nowhere)?
+                }
+                Value::String(text) => {
+                    let character = position(segment, text.chars().count())?
+                        .and_then(|i| text.chars().nth(i))
+                        .ok_or_else(nowhere)?;
+                    let index = self.characters.push(Value::String(character.to_string()));
+                    self.characters.get(index)
+                }
+                other => {
+                    return Err(Crash(format!(
+                        "'{}' object is not subscriptable",
+                        python_type_name(other)
+                    ))
+                    .into());
+                }
+            };
+        }
+        Ok(contents)
+    }
+
     /// The child `key` of a mapping node, resolved (`path / key`).
     pub fn get<'d>(&'d self, node: Node<'d>, key: &str) -> Result<Option<Node<'d>>, Crash> {
-        match node.value {
-            Value::Object(map) => match map.get(key) {
+        match pyobj::as_dict(node.value) {
+            Some(map) => match map.get(key) {
                 Some(value) => self
                     .resolve(Node {
                         value,
@@ -342,7 +411,7 @@ impl<'a> Documents<'a> {
     ) -> Result<Vec<(Key<'d>, Node<'d>)>, Crash> {
         let doc = node.doc;
         match node.value {
-            Value::Object(map) => Ok(map
+            Value::Object(map) if pyrepr::foreign(node.value).is_none() => Ok(map
                 .iter()
                 .map(|(key, value)| (Key::Str(key), Node { value, doc }))
                 .collect()),
@@ -365,67 +434,12 @@ pub(crate) enum Key<'d> {
 
 /// `is_ref` from `jsonschema-path`: a mapping whose `$ref` is a string.
 pub(crate) fn as_ref(value: &Value) -> Option<&str> {
-    value.as_object()?.get("$ref")?.as_str()
-}
-
-/// `Resource.pointer` from `referencing`.
-fn pointer<'d>(document: &'d Value, pointer: &str) -> Result<&'d Value, LookupError> {
-    let nowhere = || {
-        let mut message = format!(
-            "{} does not exist within {}",
-            pyrepr::repr_str(pointer),
-            pyrepr::repr(document)
-        );
-        if pointer == "/" {
-            message.push_str(
-                ". The pointer '/' is a valid JSON Pointer but it points to an empty string \
-                 property ''. If you intended to point to the entire resource, you should use '#'.",
-            );
-        }
-        LookupError::referencing("PointerToNowhere", message)
-    };
-    let decoded = percent_decode_str(&pointer[1..]).decode_utf8_lossy();
-    let mut contents = document;
-    for segment in decoded.split('/') {
-        contents = match contents {
-            Value::Array(items) => {
-                let index = python_int(segment).ok_or_else(|| {
-                    Crash(format!(
-                        "invalid literal for int() with base 10: {}",
-                        pyrepr::repr_str(segment)
-                    ))
-                })?;
-                let index = if index < 0 {
-                    i128::try_from(items.len()).unwrap_or(i128::MAX) + index
-                } else {
-                    index
-                };
-                usize::try_from(index)
-                    .ok()
-                    .and_then(|i| items.get(i))
-                    .ok_or_else(nowhere)?
-            }
-            Value::Object(map) => {
-                let key = segment.replace("~1", "/").replace("~0", "~");
-                map.get(&key).ok_or_else(nowhere)?
-            }
-            // Strings are sequences in Python, so indexing them yields a new
-            // one-character string that has no address in the document.
-            Value::String(_) => return Err(nowhere()),
-            other => {
-                return Err(Crash(format!(
-                    "'{}' object is not subscriptable",
-                    python_type_name(other)
-                ))
-                .into());
-            }
-        };
-    }
-    Ok(contents)
+    pyobj::as_dict(value)?.get("$ref")?.as_str()
 }
 
 /// Python `int(s)` for base-10 literals (ASCII digits, sign, underscores, whitespace).
 fn python_int(s: &str) -> Option<i128> {
+    let s = pyobj::decimal_to_ascii(s)?;
     let s = s.trim();
     let (negative, digits) = match s.as_bytes().first()? {
         b'-' => (true, &s[1..]),
@@ -526,6 +540,35 @@ mod tests {
         );
     }
 
+    #[test]
+    fn pointers_index_strings_like_python() {
+        let spec = json!({
+            "info": {"title": "tx"},
+            "a": {"$ref": "#/info/title/0"},
+            "b": {"$ref": "#/info/title/-1"},
+            "c": {"$ref": "#/info/title/x"},
+            "d": {"$ref": "#/info/title/2"},
+            "e": {"$ref": "#/info/title/1/0/0"},
+        });
+        let docs = Documents::new(&spec, "", None);
+        assert_eq!(docs.child(docs.root(), "a").unwrap().value, &json!("t"));
+        assert_eq!(docs.child(docs.root(), "b").unwrap().value, &json!("x"));
+        assert_eq!(docs.child(docs.root(), "e").unwrap().value, &json!("x"));
+        assert_eq!(
+            docs.child(docs.root(), "c").unwrap_err(),
+            Crash("invalid literal for int() with base 10: 'x'".to_owned())
+        );
+        assert!(
+            docs.child(docs.root(), "d")
+                .unwrap_err()
+                .0
+                .starts_with("'/info/title/2' does not exist within")
+        );
+        // Each resolution yields a new string, as in Python.
+        let first = docs.child(docs.root(), "a").unwrap().id();
+        assert_ne!(first, docs.child(docs.root(), "a").unwrap().id());
+    }
+
     #[test]
     fn ref_cycles_hit_the_recursion_limit() {
         let spec = json!({"a": {"$ref": "#/b"}, "b": {"$ref": "#/a"}});
@@ -553,5 +596,10 @@ mod tests {
         assert_eq!(python_int("1_0"), Some(10));
         assert_eq!(python_int("1__0"), None);
         assert_eq!(python_int("x"), None);
+        // Python's `int()` reads every script's decimal digits and whitespace.
+        assert_eq!(python_int("\u{660}\u{661}"), Some(1));
+        assert_eq!(python_int("\u{3000}-\u{ff11}\u{ff12} "), Some(-12));
+        assert_eq!(python_int("\u{660}x"), None);
+        assert_eq!(python_int("\u{2160}"), None);
     }
 }

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/crates/openapi-spec-validator-rs/src/formats.rs

@@ -682,7 +682,7 @@ fn py_int_base16(s: &str) -> Option<bool> {
 }
 
 /// Value of a Unicode decimal digit (`unicodedata.decimal`), Unicode 15.1.
-fn decimal_value(c: char) -> Option<u32> {
+pub(crate) fn decimal_value(c: char) -> Option<u32> {
     let cp = c as u32;
     let i = DECIMAL_ZEROS.partition_point(|&z| z <= cp);
     let zero = DECIMAL_ZEROS[..i].last()?;

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/crates/openapi-spec-validator-rs/src/keywords.rs

@@ -347,7 +347,7 @@ impl<'d, 'a> Walker<'d, 'a> {
             let location = self.item(parameter, "in")?;
             // jsonschema-path hands out lists and mappings as `SchemaPath`s, which are
             // hashable and equal only to the same path: such a key is never a duplicate.
-            let container = |v: &Value| v.is_array() || v.is_object();
+            let container = |v: &Value| v.is_array() || pyobj::as_dict(v).is_some();
             if container(name) || container(location) {
                 continue;
             }
@@ -451,7 +451,7 @@ impl<'d, 'a> Walker<'d, 'a> {
     /// as a nested subschema never gets the `required` check.
     fn schema(&mut self, schema: Node<'d>, require_properties: bool, meta_checked: bool) -> Step {
         let value = schema.value;
-        if !value.is_object() && !value.is_boolean() {
+        if pyobj::as_dict(value).is_none() && !value.is_boolean() {
             return self.emit_new(
                 ErrorKind::OpenApiValidation,
                 format!("{} is not of type 'object', 'boolean'", repr(value)),
@@ -675,9 +675,7 @@ impl<'d, 'a> Walker<'d, 'a> {
 
 /// `"key" in path` on a resolved node.
 fn contains(node: Node<'_>, key: &str) -> bool {
-    node.value
-        .as_object()
-        .is_some_and(|map| map.contains_key(key))
+    pyobj::as_dict(node.value).is_some_and(|map| map.contains_key(key))
 }
 
 /// `list(set(required) - set(properties) - set(nested_properties))`.

{openapi_spec_validator_rs-0.2.0 → openapi_spec_validator_rs-0.2.2}/crates/openapi-spec-validator-rs/src/lib.rs

@@ -62,7 +62,7 @@ use crate::document::Documents;
 use crate::keywords::{Stop, Walker};
 
 pub use crate::error::{ErrorKind, Outcome, SchemaErrorDetails, ValidationError, best_match};
-pub use crate::loader::{LoadError, TextErrors, load, load_text};
+pub use crate::loader::{LoadError, TextErrors, load, load_text, read_until_unreadable};
 pub use crate::pyrepr::PathItem;
 #[cfg(feature = "http")]
 pub use crate::retrieve::HttpRetriever;
@@ -188,6 +188,18 @@ pub fn validate(spec: &Value, options: &Options) -> Outcome {
     }
 }
 
+/// Reads a document file as PyYAML reads the stream `urlopen` gives upstream: only as far
+/// as the parser could get (see [`read_until_unreadable`]).
+pub(crate) fn read_file(path: &Path) -> std::io::Result<Vec<u8>> {
+    let mut file = std::fs::File::open(path)?;
+    let mut bytes = Vec::new();
+    if let Ok(metadata) = file.metadata() {
+        bytes.reserve_exact(usize::try_from(metadata.len()).unwrap_or(0));
+    }
+    read_until_unreadable(&mut file, &mut bytes)?;
+    Ok(bytes)
+}
+
 /// Stack size for validation threads: enough for documents nested as deeply as the
 /// loader allows (pages are only committed as they are used).
 const STACK_SIZE: usize = 256 << 20;
@@ -291,7 +303,7 @@ fn read_and_validate(path: &Path, options: &Options) -> Result<Outcome, LoadErro
         return Err(not_found());
     }
     let absolute = std::path::absolute(path).map_err(|_| not_found())?;
-    let bytes = std::fs::read(&absolute).map_err(|error| LoadError {
+    let bytes = read_file(&absolute).map_err(|error| LoadError {
         message: error.to_string(),
     })?;
     // Upstream reads files through `urlopen`, which is the stream name PyYAML reports.