fix: typo

Author: vadimnazarov

docs/source/concepts/basics/immutability.md

@@ -1,6 +1,6 @@
 # Pipeline-based execution: immutability and materialisation
 
-We talked a lot about data traceability or data lineage in the first two concepts. But it makes sense to implement the data lineage only if every workflow step is re-playable. Or, as we call it, reproducible. To make results reproducible, every analysis step must be easy to run again without any side effects, when the underlying data is changed.
+We talked a lot about data traceability or data lineage in the first two concepts. But it makes sense to implement the data lineage only if every workflow step is re-playable. Or, as we call it, reproducible. To make results reproducible, every analysis step must be easy to run again without any side effects (i.e., without mutating the underlying data) and, given the same inputs, producing the same outputs.
 In `immundata`-powered pipelines, an analysis is a **pipeline of immutable transformations**:
 
 * Each function returns a **new** `ImmunData` object.

docs/source/concepts/basics/immutability.zh.md

@@ -1,6 +1,6 @@
 # 基于流水线的执行：不可变性与物化（materialization）
 
-在前两个概念中我们多次谈到数据可追溯性（data lineage）。但只有当工作流程中的每一步都可被重新执行时，数据沿袭才有意义——换句话说，即**可复现**。要让结果可复现，每个分析步骤都必须能在底层数据变化时**无副作用**地被轻松重跑。
+在前两个概念中我们多次谈到数据可追溯性（data lineage）。但只有当工作流程中的每一步都可被重新执行时，数据沿袭才有意义——换句话说，即**可复现**。要让结果可复现，每个分析步骤都必须能在不改变底层数据（无副作用）的情况下被轻松重跑。
 在由 `immundata` 驱动的管线中，一次分析是由**不可变（immutable）的转换**构成的**流水线**：
 
 * 每个函数都会返回一个**全新**的 `ImmunData` 对象。