csv allow only header (#3423)

polars/polars-core/src/chunked_array/strings/json_path.rs

@@ -1,10 +1,10 @@
 use crate::prelude::*;
-use jsonpath_lib::Compiled;
+use jsonpath_lib::PathCompiled;
 use serde_json::Value;
 use std::borrow::Cow;
 
 #[cfg(feature = "extract_jsonpath")]
-fn extract_json<'a>(expr: &Compiled, json_str: &'a str) -> Option<Cow<'a, str>> {
+fn extract_json<'a>(expr: &PathCompiled, json_str: &'a str) -> Option<Cow<'a, str>> {
     serde_json::from_str(json_str).ok().and_then(|value| {
         // TODO: a lot of heap allocations here. Improve json path by adding a take?
         let result = expr.select(&value).ok()?;
@@ -23,7 +23,7 @@ impl Utf8Chunked {
     /// Refer to <https://goessner.net/articles/JsonPath/>
     #[cfg(feature = "extract_jsonpath")]
     pub fn json_path_match(&self, json_path: &str) -> Result<Utf8Chunked> {
-        match Compiled::compile(json_path) {
+        match PathCompiled::compile(json_path) {
             Ok(pat) => Ok(self.apply_on_opt(|opt_s| opt_s.and_then(|s| extract_json(&pat, s)))),
             Err(e) => Err(PolarsError::ComputeError(
                 format!("error compiling JSONpath expression {:?}", e).into(),

polars/polars-io/src/csv_core/parser.rs

@@ -89,11 +89,16 @@ where
 /// and not with
 ///     '\nfield_1,field_1'
 pub(crate) fn skip_header(input: &[u8]) -> (&[u8], usize) {
-    let mut pos = next_line_position_naive(input).expect("no lines in the file");
-    if input[pos] == b'\n' {
-        pos += 1;
+    match next_line_position_naive(input) {
+        Some(mut pos) => {
+            if input[pos] == b'\n' {
+                pos += 1;
+            }
+            (&input[pos..], pos)
+        }
+        // no lines in the file, so skipping the header is skipping all.
+        None => (&[], input.len()),
     }
-    (&input[pos..], pos)
 }
 
 /// Remove whitespace from the start of buffer.

polars/polars-io/src/csv_core/utils.rs

@@ -235,6 +235,25 @@ pub fn infer_file_schema(
             }
             column_names
         }
+    } else if has_header && !bytes.is_empty() {
+        // there was no new line char. So we copy the whole buf and add one
+        // this is likely to be cheap as there no rows.
+        let mut buf = Vec::with_capacity(bytes.len() + 2);
+        buf.extend_from_slice(bytes);
+        buf.push(b'\n');
+
+        return infer_file_schema(
+            &ReaderBytes::Owned(buf),
+            delimiter,
+            max_read_lines,
+            has_header,
+            schema_overwrite,
+            skip_rows,
+            comment_char,
+            quote_char,
+            null_values,
+            parse_dates,
+        );
     } else {
         return Err(PolarsError::NoData("empty csv".into()));
     };

polars/polars-lazy/src/frame/mod.rs

@@ -315,12 +315,15 @@ impl LazyFrame {
 
     fn rename_impl_swapping(self, mut existing: Vec<String>, mut new: Vec<String>) -> Self {
         assert_eq!(new.len(), existing.len());
-        for idx in 0..existing.len() {
-            // remove "name" -> "name"
+        let mut removed = 0;
+        for mut idx in 0..existing.len() {
+            // remove "name" -> "name
             // these are no ops.
+            idx -= removed;
             if existing[idx] == new[idx] {
                 existing.swap_remove(idx);
                 new.swap_remove(idx);
+                removed += 1;
             }
         }
 

polars/tests/it/io/csv.rs

@@ -949,8 +949,23 @@ fn test_escaping_quotes() -> Result<()> {
 fn test_header_only() -> Result<()> {
     let csv = "a,b,c";
     let file = Cursor::new(csv);
+
+    // no header
     let df = CsvReader::new(file).has_header(false).finish()?;
     assert_eq!(df.shape(), (1, 3));
+
+    // has header
+    for csv in &["x,y,z", "x,y,z\n"] {
+        let file = Cursor::new(csv);
+        let df = CsvReader::new(file).has_header(true).finish()?;
+
+        assert_eq!(df.shape(), (0, 3));
+        assert_eq!(
+            df.dtypes(),
+            &[DataType::Utf8, DataType::Utf8, DataType::Utf8]
+        );
+    }
+
     Ok(())
 }
 

py-polars/tests/test_df.py

@@ -1684,6 +1684,17 @@ def test_rename_same_name() -> None:
     df = df.rename({"groups": "groups"})
     df = df.select(["groups"])
     assert df.collect().to_dict(False) == {"groups": ["A", "A", "B", "C", "B"]}
+    df = pl.DataFrame(
+        {
+            "nrs": [1, 2, 3, 4, 5],
+            "groups": ["A", "A", "B", "C", "B"],
+            "test": [1, 2, 3, 4, 5],
+        }
+    ).lazy()
+    df = df.rename({"nrs": "nrs", "groups": "groups"})
+    df = df.select(["groups"])
+    df.collect()
+    assert df.collect().to_dict(False) == {"groups": ["A", "A", "B", "C", "B"]}
 
 
 def test_fill_null() -> None: