fix: remove dependency on regex to reduce WASM size (#29)

Using `regex`-crate increases the WASM-size of canisters by over 100 kB. This PR replaces `regex` with a custom function for "manual" parsing of `Content-Range`-header. For example the canister `http_gateway_canister_custom_assets` with `regex` had size 466'631 bytes, while with the custom function only 344'593 bytes. [TT-440](https://dfinity.atlassian.net/browse/TT-440) [TT-440]: https://dfinity.atlassian.net/browse/TT-440?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ
dfinity · Nov 12, 2024 · 1de90ba · 1de90ba
1 parent 7647269
commit 1de90ba
Show file tree

Hide file tree

Showing 6 changed files with 78 additions and 47 deletions.
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -34,7 +34,6 @@ bytes = "1"
 base64 = "0.22"
 lazy_static = "1"
 rand_chacha = "0.3"
-regex = "1"
 serde = "1"
 serde_cbor = "0.11"
 sha2 = "0.10"

diff --git a/examples/http-gateway/canister/src/custom_assets/Cargo.toml b/examples/http-gateway/canister/src/custom_assets/Cargo.toml
@@ -11,7 +11,6 @@ candid.workspace = true
 http.workspace = true
 ic-cdk.workspace = true
 ic-cdk-macros.workspace = true
-regex.workspace = true
 serde.workspace = true
 serde_cbor.workspace = true
 sha2.workspace = true

diff --git a/examples/http-gateway/canister/src/custom_assets/src/lib.rs b/examples/http-gateway/canister/src/custom_assets/src/lib.rs
@@ -162,15 +162,28 @@ fn get_header_value(headers: &[HeaderField], header_name: &str) -> Option<String
 
 fn get_content_range_begin(content_range_header_value: &str) -> usize {
     // expected format: `bytes 21010-47021/47022`
-    let re = regex::Regex::new(r"bytes\s+(\d+)-(\d+)/(\d+)").expect("invalid RE");
-    let caps = re
-        .captures(content_range_header_value)
-        .expect("malformed Content-Range header");
-    caps.get(1)
-        .expect("missing range-begin")
-        .as_str()
-        .parse()
-        .expect("malformed range-begin")
+    let str_value = content_range_header_value.trim();
+    if !str_value.starts_with("bytes ") {
+        panic!(
+            "Invalid Content-Range header: {}",
+            content_range_header_value
+        );
+    }
+    let str_value = str_value.trim_start_matches("bytes ");
+
+    let str_value_parts = str_value.split('-').collect::<Vec<_>>();
+    if str_value_parts.len() != 2 {
+        panic!(
+            "Invalid bytes spec in Content-Range header: {}",
+            content_range_header_value
+        );
+    }
+    let range_begin = str_value_parts[0].parse::<usize>().expect(&format!(
+        "Invalid range_begin in: {}",
+        content_range_header_value
+    ));
+    // Note: skipping the check whether range_end and total_length are sane.
+    range_begin
 }
 
 thread_local! {

diff --git a/packages/ic-http-gateway/Cargo.toml b/packages/ic-http-gateway/Cargo.toml
@@ -26,7 +26,6 @@ http.workspace = true
 http-body.workspace = true
 http-body-util.workspace = true
 bytes.workspace = true
-regex.workspace = true
 
 ic-agent.workspace = true
 ic-utils.workspace = true

diff --git a/packages/ic-http-gateway/src/response/response_handler.rs b/packages/ic-http-gateway/src/response/response_handler.rs
@@ -16,7 +16,6 @@ use ic_utils::{
         StreamingCallbackHttpResponse, StreamingStrategy, Token,
     },
 };
-use regex::Regex;
 
 // Limit the total number of calls to an HTTP Request loop to 1000 for now.
 static MAX_HTTP_REQUEST_STREAM_CALLBACK_CALL_COUNT: usize = 1000;
@@ -175,44 +174,68 @@ struct ContentRangeValues {
     pub total_length: usize,
 }
 
-fn parse_content_range_header_str(str_value: &str) -> Result<ContentRangeValues, AgentError> {
+fn parse_content_range_header_str(
+    content_range_str: &str,
+) -> Result<ContentRangeValues, AgentError> {
     // expected format: `bytes 21010-47021/47022`
-    let re = Regex::new(r"bytes\s+(\d+)-(\d+)/(\d+)").unwrap();
-    let Some(caps) = re.captures(str_value) else {
-        return Err(AgentError::InvalidHttpResponse(
-            "malformed Content-Range header".to_string(),
-        ));
-    };
-    let range_begin: usize = caps
-        .get(1)
-        .ok_or_else(|| AgentError::InvalidHttpResponse("missing range-begin".to_string()))?
-        .as_str()
-        .parse()
-        .map_err(|_| AgentError::InvalidHttpResponse("malformed range-begin".to_string()))?;
-    let range_end: usize = caps
-        .get(2)
-        .ok_or_else(|| AgentError::InvalidHttpResponse("missing range-end".to_string()))?
-        .as_str()
-        .parse()
-        .map_err(|_| AgentError::InvalidHttpResponse("malformed range-end".to_string()))?;
-    let total_length: usize = caps
-        .get(3)
-        .ok_or_else(|| AgentError::InvalidHttpResponse("missing size".to_string()))?
-        .as_str()
-        .parse()
-        .map_err(|_| AgentError::InvalidHttpResponse("malformed size".to_string()))?;
-    let range_values = ContentRangeValues {
+    let str_value = content_range_str.trim();
+    if !str_value.starts_with("bytes ") {
+        return Err(AgentError::InvalidHttpResponse(format!(
+            "Invalid Content-Range header '{}'",
+            content_range_str
+        )));
+    }
+    let str_value = str_value.trim_start_matches("bytes ");
+
+    let str_value_parts = str_value.split('-').collect::<Vec<_>>();
+    if str_value_parts.len() != 2 {
+        return Err(AgentError::InvalidHttpResponse(format!(
+            "Invalid bytes spec in Content-Range header '{}'",
+            content_range_str
+        )));
+    }
+    let range_begin = str_value_parts[0].parse::<usize>().map_err(|e| {
+        AgentError::InvalidHttpResponse(format!(
+            "Invalid range_begin in '{}': {}",
+            content_range_str, e
+        ))
+    })?;
+
+    let other_value_parts = str_value_parts[1].split('/').collect::<Vec<_>>();
+    if other_value_parts.len() != 2 {
+        return Err(AgentError::InvalidHttpResponse(format!(
+            "Invalid bytes spec in Content-Range header '{}'",
+            content_range_str
+        )));
+    }
+    let range_end = other_value_parts[0].parse::<usize>().map_err(|e| {
+        AgentError::InvalidHttpResponse(format!(
+            "Invalid range_end in '{}': {}",
+            content_range_str, e
+        ))
+    })?;
+    let total_length = other_value_parts[1].parse::<usize>().map_err(|e| {
+        AgentError::InvalidHttpResponse(format!(
+            "Invalid total_length in '{}': {}",
+            content_range_str, e
+        ))
+    })?;
+
+    let rv = ContentRangeValues {
         range_begin,
         range_end,
         total_length,
     };
-    if range_begin > range_end || range_begin >= total_length || range_end >= total_length {
+    if rv.range_begin > rv.range_end
+        || rv.range_begin >= rv.total_length
+        || rv.range_end >= rv.total_length
+    {
         Err(AgentError::InvalidHttpResponse(format!(
-            "inconsistent Content-Range header {:?}",
-            range_values
+            "inconsistent Content-Range header {}: {:?}",
+            content_range_str, rv
         )))
     } else {
-        Ok(range_values)
+        Ok(rv)
     }
 }
 
@@ -428,7 +451,7 @@ mod tests {
         ];
         for input in malformed_inputs {
             let result = parse_content_range_header_str(input);
-            assert_matches!(result, Err(e) if format!("{}", e).contains("malformed Content-Range header"));
+            assert_matches!(result, Err(e) if format!("{}", e).contains("Invalid "));
         }
     }
 
@@ -494,7 +517,7 @@ mod tests {
             Cow::from("bytes 42/10"),
         )];
         let result = get_initial_stream_state(http_request, canister_id, &response_headers, false);
-        assert_matches!(result, Err(e) if format!("{}", e).contains("malformed Content-Range header"));
+        assert_matches!(result, Err(e) if format!("{}", e).contains("Invalid bytes spec in Content-Range header"));
     }
 
     #[test]