From d6e5e8747e3d3196af84e25dabb16bc7583eeeae Mon Sep 17 00:00:00 2001
From: nameexhaustion <simonlin.rqmmw@slmail.me>
Date: Fri, 26 Jul 2024 22:36:35 +1000
Subject: [PATCH] fix: Scanning '%' from cloud (#17890)

---
 crates/polars-io/src/cloud/glob.rs            | 35 +++++++++++++++++--
 crates/polars-io/src/cloud/options.rs         |  6 +++-
 .../polars-io/src/path_utils/hugging_face.rs  |  7 +++-
 crates/polars-io/src/path_utils/mod.rs        | 11 +++---
 4 files changed, 51 insertions(+), 8 deletions(-)

diff --git a/crates/polars-io/src/cloud/glob.rs b/crates/polars-io/src/cloud/glob.rs
index b4d74d093a8d..12ac8081a3b8 100644
--- a/crates/polars-io/src/cloud/glob.rs
+++ b/crates/polars-io/src/cloud/glob.rs
@@ -6,7 +6,7 @@ use polars_error::PolarsResult;
 use regex::Regex;
 use url::Url;
 
-use super::CloudOptions;
+use super::{parse_url, CloudOptions};
 
 const DELIMITER: char = '/';
 
@@ -129,7 +129,7 @@ impl CloudLocation {
 
     /// Parse a CloudLocation from an url.
     pub fn new(url: &str, glob: bool) -> PolarsResult<CloudLocation> {
-        let parsed = Url::parse(url).map_err(to_compute_err)?;
+        let parsed = parse_url(url).map_err(to_compute_err)?;
         Self::from_url(&parsed, glob)
     }
 }
@@ -316,4 +316,35 @@ mod test {
             },
         )
     }
+
+    #[test]
+    fn test_cloud_location_percentages() {
+        use super::CloudLocation;
+
+        let path = "s3://bucket/%25";
+        let cloud_location = CloudLocation::new(path, true).unwrap();
+
+        assert_eq!(
+            cloud_location,
+            CloudLocation {
+                scheme: "s3".into(),
+                bucket: "bucket".into(),
+                prefix: "%25".into(),
+                expansion: None,
+            }
+        );
+
+        let path = "https://pola.rs/%25";
+        let cloud_location = CloudLocation::new(path, true).unwrap();
+
+        assert_eq!(
+            cloud_location,
+            CloudLocation {
+                scheme: "https".into(),
+                bucket: "".into(),
+                prefix: "".into(),
+                expansion: None,
+            }
+        );
+    }
 }
diff --git a/crates/polars-io/src/cloud/options.rs b/crates/polars-io/src/cloud/options.rs
index 0f48688051f7..de0968a80da0 100644
--- a/crates/polars-io/src/cloud/options.rs
+++ b/crates/polars-io/src/cloud/options.rs
@@ -154,7 +154,11 @@ impl CloudType {
 #[cfg(feature = "cloud")]
 pub(crate) fn parse_url(input: &str) -> std::result::Result<url::Url, url::ParseError> {
     Ok(if input.contains("://") {
-        url::Url::parse(input)?
+        if input.starts_with("http://") || input.starts_with("https://") {
+            url::Url::parse(input)
+        } else {
+            url::Url::parse(&input.replace("%", "%25"))
+        }?
     } else {
         let path = std::path::Path::new(input);
         let mut tmp;
diff --git a/crates/polars-io/src/path_utils/hugging_face.rs b/crates/polars-io/src/path_utils/hugging_face.rs
index cc396dc928fc..3e8f701a21ed 100644
--- a/crates/polars-io/src/path_utils/hugging_face.rs
+++ b/crates/polars-io/src/path_utils/hugging_face.rs
@@ -214,6 +214,7 @@ pub(super) async fn expand_paths_hf(
     paths: &[PathBuf],
     check_directory_level: bool,
     cloud_options: Option<&CloudOptions>,
+    glob: bool,
 ) -> PolarsResult<(usize, Vec<PathBuf>)> {
     assert!(!paths.is_empty());
 
@@ -251,7 +252,11 @@ pub(super) async fn expand_paths_hf(
         );
         let rel_path = path_parts.path.as_str();
 
-        let (prefix, expansion) = extract_prefix_expansion(rel_path)?;
+        let (prefix, expansion) = if glob {
+            extract_prefix_expansion(rel_path)?
+        } else {
+            (path_parts.path.clone(), None)
+        };
         let expansion_matcher = &if expansion.is_some() {
             Some(Matcher::new(prefix.clone(), expansion.as_deref())?)
         } else {
diff --git a/crates/polars-io/src/path_utils/mod.rs b/crates/polars-io/src/path_utils/mod.rs
index 749e248881c6..c7a751b5bd24 100644
--- a/crates/polars-io/src/path_utils/mod.rs
+++ b/crates/polars-io/src/path_utils/mod.rs
@@ -151,10 +151,13 @@ pub fn expand_paths_hive(
             use crate::cloud::object_path_from_str;
 
             if first_path.starts_with("hf://") {
-                let (expand_start_idx, paths) =
-                    crate::pl_async::get_runtime().block_on_potential_spawn(
-                        hugging_face::expand_paths_hf(paths, check_directory_level, cloud_options),
-                    )?;
+                let (expand_start_idx, paths) = crate::pl_async::get_runtime()
+                    .block_on_potential_spawn(hugging_face::expand_paths_hf(
+                        paths,
+                        check_directory_level,
+                        cloud_options,
+                        glob,
+                    ))?;
 
                 return Ok((Arc::from(paths), expand_start_idx));
             }