The workaround this user describes is exactly the kind of friction that holds spreadsheet modeling back. They are not asking for a miracle. They are asking for a fundamental capability: store a variable-length list of dates in one cell, then pass that list as a range into a lambda. That should be table stakes for any modern spreadsheet tool. Instead, they are forced into string parsing with TEXTSPLIT or hard-coded arrays on a separate sheet. Neither approach scales, and both introduce fragility where the entire point of a table is to be robust and repeatable.
What makes this situation worth pausing on is the gap it reveals between the mental model and the tool. The user already thinks in terms of functions, lookups, and dynamic arrays. They are not a novice. They understand exactly what they want the data flow to be. The problem is that the spreadsheet engine, even with its newer lambda and array features, still treats a single cell as a scalar unless the formula happens to spill. That limitation forces users to contort their data into shapes that are convenient for the software, not for the problem they are solving. The result is a solution that works, but only through cleverness and compromise.
The practical takeaway here is not that TEXTSPLIT is a bad function. It is not. It is a useful tool, and for many cases, it is the right call. But when the user has to ask whether string parsing is the *only* scalable path, they have already identified the real issue: the data model does not match the computation model. The dates are not really strings. They are a sequence. Treating them as text is a workaround, not a design choice. And every workaround adds a layer of indirection that makes the workbook harder to audit, harder to maintain, and harder for someone else to pick up later.
What this user needs is a first-class way to store and retrieve array constants within a table cell, then have lookup functions return those arrays natively. That is not an outlandish request. It is the natural next step for a spreadsheet engine that already supports LET, LAMBDA, and dynamic arrays. Until that exists, the honest answer to their question is yes, TEXTSPLIT is the pragmatic path forward, but it should not be. The fact that they had to ask whether there is a "proper" way says more about the current limits of the tool than it does about their understanding of the problem.