utils.rs

//! Handle JSON messages when writing to delta tables
use std::collections::HashMap;
use std::fmt::Display;
use std::io::Write;
use std::sync::Arc;

use crate::writer::DeltaWriterError;
use crate::DeltaTableError;

use arrow::array::{as_primitive_array, Array};
use arrow::datatypes::{
    DataType, Int16Type, Int32Type, Int64Type, Int8Type, Schema as ArrowSchema,
    SchemaRef as ArrowSchemaRef, UInt16Type, UInt32Type, UInt64Type, UInt8Type,
};
use arrow::json::reader::{Decoder, DecoderOptions};
use arrow::record_batch::*;
use object_store::path::Path;
use parking_lot::RwLock;
use serde_json::Value;
use uuid::Uuid;

const NULL_PARTITION_VALUE_DATA_PATH: &str = "__HIVE_DEFAULT_PARTITION__";

#[derive(Debug, Clone, Hash, PartialEq, Eq)]
pub(crate) struct PartitionPath {
    path: String,
}

impl PartitionPath {
    pub fn from_hashmap(
        partition_columns: &[String],
        partition_values: &HashMap<String, Option<String>>,
    ) -> Result<Self, DeltaWriterError> {
        let mut path_parts = vec![];
        for k in partition_columns.iter() {
            let partition_value = partition_values
                .get(k)
                .ok_or_else(|| DeltaWriterError::MissingPartitionColumn(k.to_string()))?;

            let partition_value = partition_value
                .as_deref()
                .unwrap_or(NULL_PARTITION_VALUE_DATA_PATH);
            let part = format!("{k}={partition_value}");

            path_parts.push(part);
        }

        Ok(PartitionPath {
            path: path_parts.join("/"),
        })
    }
}

impl From<PartitionPath> for String {
    fn from(path: PartitionPath) -> String {
        path.path
    }
}

impl AsRef<str> for PartitionPath {
    fn as_ref(&self) -> &str {
        &self.path
    }
}

impl Display for PartitionPath {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> Result<(), std::fmt::Error> {
        self.path.fmt(f)
    }
}

// TODO: parquet files have a 5 digit zero-padded prefix and a "c\d{3}" suffix that
// I have not been able to find documentation for yet.
pub(crate) fn next_data_path(
    partition_columns: &[String],
    partition_values: &HashMap<String, Option<String>>,
    part: Option<i32>,
) -> Result<Path, DeltaWriterError> {
    // TODO: what does 00000 mean?
    // TODO (roeap): my understanding is, that the values are used as a counter - i.e. if a single batch of
    // data written to one partition needs to be split due to desired file size constraints.
    let first_part = match part {
        Some(count) => format!("{count:0>5}"),
        _ => "00000".to_string(),
    };
    let uuid_part = Uuid::new_v4();
    // TODO: what does c000 mean?
    let last_part = "c000";

    // NOTE: If we add a non-snappy option, file name must change
    let file_name = format!("part-{first_part}-{uuid_part}-{last_part}.snappy.parquet");

    if partition_columns.is_empty() {
        return Ok(Path::from(file_name));
    }

    let partition_key = PartitionPath::from_hashmap(partition_columns, partition_values)?;
    Ok(Path::from(format!("{partition_key}/{file_name}")))
}

/// Convert a vector of json values to a RecordBatch
pub fn record_batch_from_message(
    arrow_schema: Arc<ArrowSchema>,
    message_buffer: &[Value],
) -> Result<RecordBatch, DeltaTableError> {
    let mut value_iter = message_buffer.iter().map(|j| Ok(j.to_owned()));
    let options = DecoderOptions::new().with_batch_size(message_buffer.len());
    let decoder = Decoder::new(arrow_schema, options);
    decoder
        .next_batch(&mut value_iter)?
        .ok_or_else(|| DeltaWriterError::EmptyRecordBatch.into())
}

// very naive implementation for plucking the partition value from the first element of a column array.
// ideally, we would do some validation to ensure the record batch containing the passed partition column contains only distinct values.
// if we calculate stats _first_, we can avoid the extra iteration by ensuring max and min match for the column.
// however, stats are optional and can be added later with `dataChange` false log entries, and it may be more appropriate to add stats _later_ to speed up the initial write.
// a happy middle-road might be to compute stats for partition columns only on the initial write since we should validate partition values anyway, and compute additional stats later (at checkpoint time perhaps?).
// also this does not currently support nested partition columns and many other data types.
// TODO is this comment still valid, since we should be sure now, that the arrays where this
// gets aplied have a single unique value
pub(crate) fn stringified_partition_value(
    arr: &Arc<dyn Array>,
) -> Result<Option<String>, DeltaWriterError> {
    let data_type = arr.data_type();

    if arr.is_null(0) {
        return Ok(None);
    }

    let s = match data_type {
        DataType::Int8 => as_primitive_array::<Int8Type>(arr).value(0).to_string(),
        DataType::Int16 => as_primitive_array::<Int16Type>(arr).value(0).to_string(),
        DataType::Int32 => as_primitive_array::<Int32Type>(arr).value(0).to_string(),
        DataType::Int64 => as_primitive_array::<Int64Type>(arr).value(0).to_string(),
        DataType::UInt8 => as_primitive_array::<UInt8Type>(arr).value(0).to_string(),
        DataType::UInt16 => as_primitive_array::<UInt16Type>(arr).value(0).to_string(),
        DataType::UInt32 => as_primitive_array::<UInt32Type>(arr).value(0).to_string(),
        DataType::UInt64 => as_primitive_array::<UInt64Type>(arr).value(0).to_string(),
        DataType::Utf8 => {
            let data = arrow::array::as_string_array(arr);

            data.value(0).to_string()
        }
        // TODO: handle more types
        _ => {
            unimplemented!("Unimplemented data type: {:?}", data_type);
        }
    };

    Ok(Some(s))
}

/// Remove any partition related columns from the record batch
pub(crate) fn record_batch_without_partitions(
    record_batch: &RecordBatch,
    partition_columns: &[String],
) -> Result<RecordBatch, DeltaWriterError> {
    let mut non_partition_columns = Vec::new();
    for (i, field) in record_batch.schema().fields().iter().enumerate() {
        if !partition_columns.contains(field.name()) {
            non_partition_columns.push(i);
        }
    }

    Ok(record_batch.project(&non_partition_columns)?)
}

/// Arrow schema for the physical file which has partition columns removed
pub(crate) fn arrow_schema_without_partitions(
    arrow_schema: &Arc<ArrowSchema>,
    partition_columns: &[String],
) -> ArrowSchemaRef {
    Arc::new(ArrowSchema::new(
        arrow_schema
            .fields()
            .iter()
            .filter(|f| !partition_columns.contains(f.name()))
            .map(|f| f.to_owned())
            .collect::<Vec<_>>(),
    ))
}

/// An in memory buffer that allows for shared ownership and interior mutability.
/// The underlying buffer is wrapped in an `Arc` and `RwLock`, so cloning the instance
/// allows multiple owners to have access to the same underlying buffer.
#[derive(Debug, Default, Clone)]
pub struct ShareableBuffer {
    buffer: Arc<RwLock<Vec<u8>>>,
}

impl ShareableBuffer {
    /// Consumes this instance and returns the underlying buffer.
    /// Returns None if there are other references to the instance.
    pub fn into_inner(self) -> Option<Vec<u8>> {
        Arc::try_unwrap(self.buffer)
            .ok()
            .map(|lock| lock.into_inner())
    }

    /// Returns a clone of the the underlying buffer as a `Vec`.
    pub fn to_vec(&self) -> Vec<u8> {
        let inner = self.buffer.read();
        (*inner).to_vec()
    }

    /// Returns the number of bytes in the underlying buffer.
    pub fn len(&self) -> usize {
        let inner = self.buffer.read();
        (*inner).len()
    }

    /// Returns true if the underlying buffer is empty.
    pub fn is_empty(&self) -> bool {
        let inner = self.buffer.read();
        (*inner).is_empty()
    }

    /// Creates a new instance with buffer initialized from the underylying bytes.
    pub fn from_bytes(bytes: &[u8]) -> Self {
        Self {
            buffer: Arc::new(RwLock::new(bytes.to_vec())),
        }
    }
}

impl Write for ShareableBuffer {
    fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
        let mut inner = self.buffer.write();
        (*inner).write(buf)
    }

    fn flush(&mut self) -> std::io::Result<()> {
        let mut inner = self.buffer.write();
        (*inner).flush()
    }
}