Fix the issues of strip and merge desc for code block

Author: chenyukang

schemars/tests/docs.rs

@@ -54,6 +54,59 @@ enum MyEnum {
     },
 }
 
+#[allow(dead_code)]
+#[derive(JsonSchema)]
+/**
+ *
+ * # This is the struct's title
+ *
+ * This is the struct's description.
+ *
+ * This is example:
+ * ```json
+ * {
+ *    "value": 0,
+ *    "type": "msg"
+ * }
+ * ```
+ */
+struct MyStructWithInlineCode {
+    /// # An integer
+    my_int: i32,
+}
+
+#[allow(dead_code)]
+#[derive(JsonSchema)]
+
+/// This is example:
+/// ```json
+/// {
+///    "value": 0,
+///    "type": "msg"
+/// }
+/// ```
+struct MyStructWithInlineCodeNormal {
+    /// # An integer
+    my_int: i32,
+}
+
+#[allow(dead_code)]
+#[derive(JsonSchema)]
+
+/// This is example:
+///
+/// | A | B |
+/// |---|---|
+/// | 1 | 2 |
+/// | 3 | 4 |
+/// | 5 | 6 |
+/// this is last line
+///
+struct MyStructWithInlineCodeTable {
+    /// # An integer
+    my_int: i32,
+}
+
 #[test]
 fn doc_comments_struct() -> TestResult {
     test_default_generated_schema::<MyStruct>("doc_comments_struct")
@@ -70,6 +123,23 @@ fn doc_comments_enum() -> TestResult {
     test_default_generated_schema::<MyEnum>("doc_comments_enum")
 }
 
+#[test]
+fn doc_comments_with_inline_code() -> TestResult {
+    test_default_generated_schema::<MyStructWithInlineCode>("doc_comments_with_inline_code")
+}
+
+#[test]
+fn doc_comments_with_inline_code_normal() -> TestResult {
+    test_default_generated_schema::<MyStructWithInlineCodeNormal>(
+        "doc_comments_with_inline_code_normal",
+    )
+}
+
+#[test]
+fn doc_comments_with_inline_table() -> TestResult {
+    test_default_generated_schema::<MyStructWithInlineCodeTable>("doc_comments_with_inline_table")
+}
+
 /// # OverrideDocs struct
 /// This description should be overridden
 #[allow(dead_code)]

schemars/tests/expected/doc_comments_with_inline_code.json

@@ -0,0 +1,16 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "This is the struct's title",
+  "description": "This is the struct's description.\n\nThis is example:\n ```json\n {\n    \"value\": 0,\n    \"type\": \"msg\"\n }\n```",
+  "type": "object",
+  "required": [
+    "my_int"
+  ],
+  "properties": {
+    "my_int": {
+      "title": "An integer",
+      "type": "integer",
+      "format": "int32"
+    }
+  }
+}

schemars/tests/expected/doc_comments_with_inline_code_normal.json

@@ -0,0 +1,16 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "MyStructWithInlineCodeNormal",
+  "description": "This is example:\n ```json\n {\n    \"value\": 0,\n    \"type\": \"msg\"\n }\n```",
+  "type": "object",
+  "required": [
+    "my_int"
+  ],
+  "properties": {
+    "my_int": {
+      "title": "An integer",
+      "type": "integer",
+      "format": "int32"
+    }
+  }
+}

schemars/tests/expected/doc_comments_with_inline_table.json

@@ -0,0 +1,16 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "MyStructWithInlineCodeTable",
+  "description": "This is example:\n\n| A | B |\n|---|---|\n| 1 | 2 |\n| 3 | 4 |\n| 5 | 6 | this is last line",
+  "type": "object",
+  "required": [
+    "my_int"
+  ],
+  "properties": {
+    "my_int": {
+      "title": "An integer",
+      "type": "integer",
+      "format": "int32"
+    }
+  }
+}

schemars/tests/validate_inner.rs

@@ -12,7 +12,7 @@ pub struct Struct<'a> {
     #[schemars(inner(length(min = 5, max = 100)))]
     array_str_length: [&'a str; 2],
     #[schemars(inner(contains(pattern = "substring...")))]
-    slice_str_contains: &'a[&'a str],
+    slice_str_contains: &'a [&'a str],
     #[schemars(inner(regex = "STARTS_WITH_HELLO"))]
     vec_str_regex: Vec<String>,
     #[schemars(inner(length(min = 1, max = 100)))]

schemars_derive/src/attr/doc.rs

@@ -25,7 +25,13 @@ fn merge_description_lines(doc: &str) -> Option<String> {
     let desc = doc
         .trim()
         .split("\n\n")
-        .filter_map(|line| none_if_empty(line.trim().replace('\n', " ")))
+        .map(|s| s.to_string())
+        .collect::<Vec<_>>();
+
+    let desc = desc
+        .iter()
+        .map(|paragrah| merge_without_codeblock(paragrah))
+        .filter_map(|line| none_if_empty(line.to_string()))
         .collect::<Vec<_>>()
         .join("\n\n");
     none_if_empty(desc)
@@ -47,30 +53,22 @@ fn get_doc(attrs: &[Attribute]) -> Option<String> {
             None
         })
         .collect::<Vec<_>>();
-
-    let mut lines = attrs
+    let lines = attrs
         .iter()
         .flat_map(|a| a.split('\n'))
-        .map(str::trim)
         .skip_while(|s| s.is_empty())
+        .map(|l| l.to_string())
         .collect::<Vec<_>>();
 
-    if let Some(&"") = lines.last() {
-        lines.pop();
-    }
-
+    let mut res = strip_without_codeblock(&lines, |l| l.trim().to_string());
     // Added for backward-compatibility, but perhaps we shouldn't do this
     // https://github.com/rust-lang/rust/issues/32088
-    if lines.iter().all(|l| l.starts_with('*')) {
-        for line in lines.iter_mut() {
-            *line = line[1..].trim()
-        }
-        while let Some(&"") = lines.first() {
-            lines.remove(0);
-        }
+    if res.iter().all(|l| l.trim().starts_with('*')) {
+        res = res.iter().map(|l| l[1..].to_string()).collect::<Vec<_>>();
+        res = strip_without_codeblock(&res, |l| l.trim().to_string());
     };
 
-    none_if_empty(lines.join("\n"))
+    none_if_empty(res.join("\n"))
 }
 
 fn none_if_empty(s: String) -> Option<String> {
@@ -80,3 +78,50 @@ fn none_if_empty(s: String) -> Option<String> {
         Some(s)
     }
 }
+
+fn strip_without_codeblock(lines: &Vec<String>, func: fn(&str) -> String) -> Vec<String> {
+    let mut res = vec![];
+    let mut in_codeblock = false;
+    for line in lines {
+        if line.trim().starts_with("```") {
+            in_codeblock = !in_codeblock;
+        }
+        let l = if in_codeblock {
+            line.to_string()
+        } else {
+            func(line)
+        };
+        res.push(l);
+    }
+    while let Some("") = res.first().map(|s| s.as_str()) {
+        res.remove(0);
+    }
+    while let Some("") = res.last().map(|s| s.as_str()) {
+        res.pop();
+    }
+    res
+}
+
+fn merge_without_codeblock(content: &str) -> String {
+    let lines = content.lines();
+    let mut res = String::new();
+    let mut in_codeblock = false;
+    for line in lines {
+        let flag = line.trim().starts_with("```");
+        if flag {
+            in_codeblock = !in_codeblock;
+        }
+        // other possible Markdown prefix characters
+        let maybe_markdown = ["#", "-", ">", "|", "*", "["]
+            .iter()
+            .any(|p| line.trim().starts_with(p))
+            || line.trim().chars().next().map(char::is_numeric) == Some(true);
+        let prefix = if in_codeblock || flag || maybe_markdown {
+            "\n"
+        } else {
+            " "
+        };
+        res += &(format!("{}{}", prefix, line));
+    }
+    res.trim().to_string()
+}