#62 cli tool for printing file stats (#84)

* #62 cli tool for printing file stats

* Update Cargo.toml

Co-authored-by: Weny Xu <wenymedia@gmail.com>

* fixed formatting

---------

Co-authored-by: Weny Xu <wenymedia@gmail.com>

Cargo.toml

@@ -12,11 +12,12 @@ include = ["src/**/*.rs", "Cargo.toml"]
 rust-version = "1.70"
 
 [dependencies]
+anyhow = { version = "1.0", optional = true }
 arrow = { version = "51", features = ["prettyprint"] }
 bytes = "1.4"
 chrono = { version = "0.4.37", default-features = false, features = ["std"] }
 chrono-tz = "0.8.6"
-clap = { version = "4.5.3", features = ["derive"], optional = true }
+clap = { version = "4.5.4", features = ["derive"], optional = true }
 fallible-streaming-iterator = { version = "0.1" }
 flate2 = "1"
 futures = { version = "0.3", optional = true, default-features = false, features = ["std"] }
@@ -52,7 +53,7 @@ serde_json = { version = "1.0", default-features = false, features = ["std"] }
 default = ["async"]
 
 async = ["futures", "futures-util", "tokio"]
-cli = ["clap"]
+cli = ["anyhow", "clap"]
 
 [[bench]]
 name = "arrow_reader"
@@ -70,3 +71,7 @@ path = "./examples/datafusion_integration.rs"
 [[bin]]
 name = "orc-metadata"
 required-features = ["cli"]
+
+[[bin]]
+name = "orc-stats"
+required-features = ["cli"]

src/bin/orc-stats.rs

@@ -0,0 +1,132 @@
+use std::{fs::File, path::PathBuf, sync::Arc};
+
+use anyhow::Result;
+use arrow::temporal_conversions::{date32_to_datetime, timestamp_ms_to_datetime};
+use clap::Parser;
+use orc_rust::{reader::metadata::read_metadata, statistics::ColumnStatistics};
+
+#[derive(Parser)]
+#[command(name = "orc-stats")]
+#[command(version, about = "Print column and stripe stats from the orc file", long_about = None)]
+struct Cli {
+    /// Path to the orc file
+    file: PathBuf,
+}
+
+fn print_column_stats(col_stats: &ColumnStatistics) {
+    if let Some(tstats) = col_stats.type_statistics() {
+        match tstats {
+            orc_rust::statistics::TypeStatistics::Integer { min, max, sum } => {
+                println!("* Data type Integer");
+                println!("* Minimum: {}", min);
+                println!("* Maximum: {}", max);
+                if let Some(sum) = sum {
+                    println!("* Sum: {}", sum);
+                }
+            }
+            orc_rust::statistics::TypeStatistics::Double { min, max, sum } => {
+                println!("* Data type Double");
+                println!("* Minimum: {}", min);
+                println!("* Maximum: {}", max);
+                if let Some(sum) = sum {
+                    println!("* Sum: {}", sum);
+                }
+            }
+            orc_rust::statistics::TypeStatistics::String { min, max, sum } => {
+                println!("* Data type String");
+                println!("* Minimum: {}", min);
+                println!("* Maximum: {}", max);
+                println!("* Sum: {}", sum);
+            }
+            orc_rust::statistics::TypeStatistics::Bucket { true_count } => {
+                println!("* Data type Bucket");
+                println!("* True count: {}", true_count);
+            }
+            orc_rust::statistics::TypeStatistics::Decimal { min, max, sum } => {
+                println!("* Data type Decimal");
+                println!("* Minimum: {}", min);
+                println!("* Maximum: {}", max);
+                println!("* Sum: {}", sum);
+            }
+            orc_rust::statistics::TypeStatistics::Date { min, max } => {
+                println!("* Data type Date");
+                if let Some(dt) = date32_to_datetime(*min) {
+                    println!("* Minimum: {}", dt);
+                }
+                if let Some(dt) = date32_to_datetime(*max) {
+                    println!("* Maximum: {}", dt);
+                }
+            }
+            orc_rust::statistics::TypeStatistics::Binary { sum } => {
+                println!("* Data type Binary");
+                println!("* Sum: {}", sum);
+            }
+            orc_rust::statistics::TypeStatistics::Timestamp {
+                min,
+                max,
+                min_utc,
+                max_utc,
+            } => {
+                println!("* Data type Timestamp");
+                println!("* Minimum: {}", min);
+                println!("* Maximum: {}", max);
+                if let Some(ts) = timestamp_ms_to_datetime(*min_utc) {
+                    println!("* Minimum UTC: {}", ts);
+                }
+                if let Some(ts) = timestamp_ms_to_datetime(*max_utc) {
+                    println!("* Maximum UTC: {}", ts);
+                }
+            }
+            orc_rust::statistics::TypeStatistics::Collection {
+                min_children,
+                max_children,
+                total_children,
+            } => {
+                println!("* Data type Collection");
+                println!("* Minimum children: {}", min_children);
+                println!("* Maximum children: {}", max_children);
+                println!("* Total children: {}", total_children);
+            }
+        }
+    }
+
+    println!("* Num values: {}", col_stats.number_of_values());
+    println!("* Has nulls: {}", col_stats.has_null());
+    println!("");
+}
+
+fn main() -> Result<()> {
+    let cli = Cli::parse();
+
+    let mut f = File::open(&cli.file)?;
+    let metadata = Arc::new(read_metadata(&mut f)?);
+
+    println!("# Column stats");
+    println!(
+        "File {:?} has {} columns",
+        cli.file,
+        metadata.column_file_statistics().len()
+    );
+    println!();
+    for (idx, col_stats) in metadata.column_file_statistics().iter().enumerate() {
+        println!("## Column {idx}");
+        print_column_stats(col_stats);
+    }
+
+    println!("# Stripe stats");
+    println!(
+        "File {:?} has {} stripes",
+        cli.file,
+        metadata.stripe_metadatas().len()
+    );
+    println!();
+    for (idm, sm) in metadata.stripe_metadatas().iter().enumerate() {
+        println!("----- Stripe {idm} -----\n");
+        for (idc, col_stats) in sm.column_statistics().iter().enumerate() {
+            println!("## Column {idc}");
+            print_column_stats(col_stats);
+        }
+    }
+
+    Ok(())
+}