From 842ff4c15ea988b9a8b99c05ca17894e7ad1f1d6 Mon Sep 17 00:00:00 2001
From: Kenny Daniel <platypii@gmail.com>
Date: Mon, 21 Apr 2025 22:47:43 -0700
Subject: [PATCH] Repeated types not yet supported

---
 src/column.js    | 14 +++++++-------
 src/datapage.js  | 12 +++++++-----
 src/unconvert.js |  2 +-
 3 files changed, 15 insertions(+), 13 deletions(-)

diff --git a/src/column.js b/src/column.js
index 8006fd1..a781290 100644
--- a/src/column.js
+++ b/src/column.js
@@ -13,9 +13,9 @@ import { writeDataPageV2, writePageHeader } from './datapage.js'
  * @returns {ColumnMetaData}
  */
 export function writeColumn(writer, schemaPath, values, compressed, stats) {
-  const schemaElement = schemaPath[schemaPath.length - 1]
-  const { type, type_length } = schemaElement
-  if (!type) throw new Error(`column ${schemaElement.name} cannot determine type`)
+  const element = schemaPath[schemaPath.length - 1]
+  const { type, type_length } = element
+  if (!type) throw new Error(`column ${element.name} cannot determine type`)
   const offsetStart = writer.offset
   const num_values = values.length
   /** @type {Encoding[]} */
@@ -41,20 +41,20 @@ export function writeColumn(writer, schemaPath, values, compressed, stats) {
     }
 
     // write unconverted dictionary page
-    const unconverted = unconvert(schemaElement, dictionary)
+    const unconverted = unconvert(element, dictionary)
     writeDictionaryPage(writer, unconverted, type, type_length, compressed)
 
     // write data page with dictionary indexes
     data_page_offset = BigInt(writer.offset)
-    writeDataPageV2(writer, indexes, type, schemaPath, 'RLE_DICTIONARY', compressed)
+    writeDataPageV2(writer, indexes, schemaPath, 'RLE_DICTIONARY', compressed)
     encodings.push('RLE_DICTIONARY')
   } else {
     // unconvert values from rich types to simple
-    values = unconvert(schemaElement, values)
+    values = unconvert(element, values)
 
     // write data page
     const encoding = type === 'BOOLEAN' && values.length > 16 ? 'RLE' : 'PLAIN'
-    writeDataPageV2(writer, values, type, schemaPath, encoding, compressed)
+    writeDataPageV2(writer, values, schemaPath, encoding, compressed)
     encodings.push(encoding)
   }
 
diff --git a/src/datapage.js b/src/datapage.js
index df4883e..bb4704d 100644
--- a/src/datapage.js
+++ b/src/datapage.js
@@ -10,13 +10,15 @@ import { getMaxDefinitionLevel, getMaxRepetitionLevel } from './schema.js'
  * @import {Writer} from '../src/types.js'
  * @param {Writer} writer
  * @param {DecodedArray} values
- * @param {ParquetType} type
  * @param {SchemaElement[]} schemaPath
  * @param {import('hyparquet').Encoding} encoding
  * @param {boolean} compressed
  */
-export function writeDataPageV2(writer, values, type, schemaPath, encoding, compressed) {
-  const fixedLength = schemaPath.at(-1)?.type_length
+export function writeDataPageV2(writer, values, schemaPath, encoding, compressed) {
+  const { name, type, type_length, repetition_type } = schemaPath[schemaPath.length - 1]
+
+  if (!type) throw new Error(`column ${name} cannot determine type`)
+  if (repetition_type === 'REPEATED') throw new Error(`column ${name} repeated types not supported`)
 
   // write levels to temp buffer
   const levels = new ByteWriter()
@@ -39,7 +41,7 @@ export function writeDataPageV2(writer, values, type, schemaPath, encoding, comp
     page.appendUint8(bitWidth) // prepend bitWidth
     writeRleBitPackedHybrid(page, nonnull, bitWidth)
   } else {
-    writePlain(page, nonnull, type, fixedLength)
+    writePlain(page, nonnull, type, type_length)
   }
 
   // compress page data
@@ -108,7 +110,7 @@ export function writePageHeader(writer, header) {
 }
 
 /**
- * @import {DecodedArray, PageHeader, ParquetType, SchemaElement} from 'hyparquet'
+ * @import {DecodedArray, PageHeader, SchemaElement} from 'hyparquet'
  * @param {Writer} writer
  * @param {SchemaElement[]} schemaPath
  * @param {DecodedArray} values
diff --git a/src/unconvert.js b/src/unconvert.js
index c0c6bf0..839e98b 100644
--- a/src/unconvert.js
+++ b/src/unconvert.js
@@ -35,7 +35,7 @@ export function unconvert(element, values) {
   if (ctype === 'UTF8') {
     if (!Array.isArray(values)) throw new Error('strings must be an array')
     const encoder = new TextEncoder()
-    return values.map(v => encoder.encode(v))
+    return values.map(v => typeof v === 'string' ? encoder.encode(v) : v)
   }
   if (ltype?.type === 'FLOAT16') {
     if (type !== 'FIXED_LEN_BYTE_ARRAY') throw new Error('FLOAT16 must be FIXED_LEN_BYTE_ARRAY type')