Hacktoberfest 2026:メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

[Bug]: MCP orderby takes incompatible shapes in read_records and aggregate_records, and the mismatch returns an opaque UnexpectedError

クローズ
#3,810 コメント 1 件 リアクション 0 件 担当者 1 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

@aaronburtle がすでに取り組んでいます。

2026年9月14日 から。

評価

この issue はまだ評価されていません。

説明

mcp-server
What happened?

The two MCP DML tools accept orderby in mutually incompatible shapes. Passing one tool's shape to the other returns UnexpectedError with no indication of what was wrong, so the caller has no way to discover the correct form from the error.

aggregate_records expects a bare direction string. It sorts by the aggregated value.

argument result
orderby: "desc" OK, groups sorted by aggregated value descending
orderby: "asc" OK
orderby: "COLUMN_NAME" InvalidArguments: Argument 'orderby' must be either 'asc' or 'desc' when provided. Got: 'COLUMN_NAME'. — clear and actionable
orderby: ["desc"] UnexpectedError: Unexpected error occurred in AggregateRecordsTool. — opaque

read_records expects an array of column specifications.

argument result
orderby: ["COLUMN_NAME desc"] OK
orderby: "COLUMN_NAME desc" UnexpectedError: Unexpected error occurred in ReadRecordsTool. — opaque
orderby: "COLUMN_NAME" same

So the shape that is correct in one tool is an opaque failure in the other, in both directions.

Why this matters more for MCP than for REST

The consumer is a language model reading describe_entities output and the tool input schemas. Having learned ["COLUMN desc"] from read_records, it will use that in aggregate_records and get a failure that names neither the parameter nor the expected form. There is no discovery path from the error back to the working call.

Asks
  1. Return an actionable message for the shape mismatch in both tools, as aggregate_records already does for the column-name case. Naming the expected shape would be enough.
  2. Consider accepting both shapes, or aligning them.
Two related observations

aggregate_records silently ignores orderby when groupby is absent, rather than rejecting it. A caller asking for sorted output gets unsorted output and a success response.

Grouped results are returned ordered by the aggregated value, and there is no way to order by the group key, because orderby accepts only a direction. For a groupby on a date or period column this means the caller cannot ask for chronological order and has to sort client-side. Given that filtering on a date-typed column is currently broken (#3760), groupby on the date is the only way to reach a single period, which makes the inability to order by that key more limiting than it would otherwise be.

Version

2.1.3-rc

What database are you using?

Azure SQL

What hosting model are you using?

Azure Container Apps

Which API approach are you accessing DAB through?

MCP

主要言語
C#
スター
1.5k
フォーク
372
平均マージ
9日 2時間
マージ済み PR(30日)
10

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

Azure/data-api-builder のほかの issue

Azure/data-api-builder の issue をすべて見る

似ている issue

C# の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。