본문으로 건너뛰기

ADR 0045 - Function batch — rejected per ADR-0043

  • Status: rejected
  • Date: 2026-05-18
  • Spec target: XTL 0.1 (rejection; no spec change)

Context​

ADR-0043 settled the Excel-native preference principle: a new XTL function is warranted only when its evaluation must happen before rendering. ADR-0044 applied the principle to accept 6 functions. This ADR documents the rejection of the remaining Tier-1 batch candidates so the questions stay settled.

Each rejected proposal below was surfaced as a "JXLS / Excel has this — should xl3?" question during the gap analysis after batch 2 (ADR-0038..0042). Each fails ADR-0043's test: the same value can be produced by an Excel formula in the output cell.

Rejected proposals​

Math expansion: SQRT, POWER, MOD, INT, CEILING, FLOOR, SIGN​

Why rejected:

  • Render-time uses are rare. Filters / sorts almost never need SQRT([X]) or POWER([X], 2).
  • The output-cell use is trivially served by an Excel formula: =SQRT(B2), =POWER(B2, 2), =MOD(B2, 7), =INT(B2).
  • xl3 already exposes ROUND and ABS because both surface in filter predicates. The math expansion proposals do not.

Type tests: ISBLANK, ISNUMBER, ISTEXT, ISDATE, ISERROR​

Why rejected:

  • IFEMPTY(value, fallback) already covers the blank-fallback pattern (the dominant render-time use).
  • IFERROR(value, fallback) (added by ADR-0044) covers the error-fallback pattern.
  • For cell-output type checks the author writes =ISNUMBER(B2) / =ISBLANK(B2) and Excel evaluates.
  • Type-based filter predicates exist but are exotic — when they appear the author can compare the result of an existing function (e.g., LEN(TRIM([X])) = 0 for "blank-ish").

NOW, WEEKDAY, WEEKNUM, NETWORKDAYS​

Why rejected:

  • TODAY() covers the "render-time UTC date" case. A render-time timestamp with sub-day precision (NOW) is rarely meaningful in operations reports — the filename pattern is the only place it might surface and minute-granular filenames are a footgun.
  • WEEKDAY / WEEKNUM / NETWORKDAYS are all Excel-native formulas. For cell output: =WEEKDAY(B2) works. For filters they are exotic.
  • xl3's DATEDIF(..., "D") (per ADR-0019 amendment) handles the "days between" question with the same UTC precision.

Conditional aggregates: SUMIF, COUNTIF, AVERAGEIF, SUMIFS​

Why rejected (narrow claim):

  • For the block-total pattern — "render rows matching some criteria and a total at the bottom" — xl3's @filter + SUM combination is equivalent. The author writes: 
    {{ @filter [Status] = "VIP" }}
    ... data block ...
    {{ SUM([Amount]) }}
    in place of =SUMIFS(Amount, Status, "VIP") and gets the same computed total.
  • For the cross-cutting pattern — "render the full set of rows AND show a conditional total in a separate cell" — @filter is not equivalent (it would remove the rows from the output). The author's path is to write the Excel formula directly in the output cell: =SUMIF(B:B, "VIP", C:C). xl3 preserves the formula and Excel evaluates at open.
  • The block-total path is the dominant case; the cross-cutting case is rare and has a clean Excel-formula answer. Neither case justifies adding XTL SUMIF.

TEXT() format token expansion: currency ₩, percent %,​

accounting parens, scientific E+00, conditional color​

Why rejected (with one acknowledged gap):

  • This is the case that motivated ADR-0043 in the first place. Visual formatting belongs on the cell's numFmt, not in a string-returning function.
  • Authors who want ₩1,200,000 in the output write the cell with numFmt "₩"#,##0 and substitute the number.
  • Authors who want the formatted string in concatenation ("Total: " & TEXT([Amount], "#,##0")) still get the current XTL-0.1 token set, which covers the common case.

Acknowledged gap — filename composition:

output_file_pattern is a string, not a cell — numFmt has no target there. An author who wants output-12345-USD.xlsx cannot get currency-formatted amount in the filename via the rejected TEXT() expansion. The temporary workaround is manual concatenation: "USD-" & TEXT([Amount], "#,##0") or upstream preparation in __inputs__. This is a legitimately deferred gap under ADR-0043 Category 3 (render-time scalar that flows back into XTL context). A re-proposal limited to filename composition is the path forward when demand surfaces.

SUMPRODUCT, MEDIAN, STDEV, PERCENTILE, NPV, PMT, FV​

Why rejected:

  • Statistical and financial functions are pure Excel-formula territory. They operate on cell ranges in the output workbook and have no render-time variant.
  • A user with a =SUMPRODUCT(B2:B100, C2:C100) cell in their template gets the right behavior automatically — the formula is preserved verbatim and Excel evaluates at open.

Why rejected:

  • Excel-native for cell output: =FIND("@", B2).
  • Render-time use case (find substring in __inputs__) is exotic and can be done with the existing TEXT(IF(... LIKE pattern), ...) patterns once string concatenation is in play.

Number constructors: VALUE(s)​

Why rejected:

  • xl3's toNumber coercion in arithmetic already does this implicitly. [StringNumber] * 1 or [StringNumber] + 0 is the XTL idiom.
  • For cell output, =VALUE(B2) is Excel-native.

How re-proposals should work​

This ADR is not absolute. Re-proposing any rejected function is allowed under ADR-0034 § "Disagreements" and ADR-0043's principle: the re-proposal must demonstrate at least one of the five render-time-critical categories from ADR-0043's "How to apply" section. A "this is convenient" argument alone is not sufficient.

The clearest path to re-acceptance: a real-world template fails because an Excel formula won't work in that position (e.g., the function needs to feed a @filter predicate). A fixture or issue illustrating the case is the strongest evidence.

Consequences​

  • Future "should xl3 have X?" questions for any function in this list have a documented answer to cite.
  • Porters know the surface is intentionally smaller than Excel's full catalog.
  • The cookbook recipe ADR-0050 references this ADR so authors hitting a rejected function know where to look (the cell-formula path).

References​

  • ADR-0043 — Excel-native preference principle (the gate)
  • ADR-0044 — function batch accepted
  • ADR-0034 — Relationship to prior-art template engines (Corollary 3 spirit)
  • ADR-0037, ADR-0042 — prior rejection ADRs (format precedent)