Amiga GFA-BASIC 3.5 interpreter addendum — extracted feature inventory
Amiga GFA-BASIC 3.5 interpreter addendum — extracted feature inventory
Primary source: the supplied GFA_Basic_3.5_Interpreter.pdf; independently cross-checked against GFA-Basic 3.0 + 3.5 Upd (Amiga).pdf (487 PDF pages). The second scan's PDF p. 3 identifies the base as GFA BASIC 3.0 Interpreter; p. 4 gives ©1988 GFA Systemtechnik GmbH, program author Frank Ostrowski, second edition December 1988; p. 7 explicitly instructs readers to add Chapter 13, Chapter 14, and the 3.5 editor changes to the 3.0 manual. The following references are printed addendum page numbers.
1. Editor changes (printed base-manual p. 1-25; explicit 3.5 update)
- Editor uses two more bytes per program row than versions 3.0–3.041; the source says this makes backward scrolling faster and enables folding functions.
- Search can also find closed procedure/function header rows.
- During listing, labels are shifted two characters to the left, as with
CASE.
2. Tab-key operations (printed base-manual p. 1-25; explicit 3.5 update)
| Key | Documented action |
|---|---|
| Tab | Move cursor to next tabulator position |
| Ctrl+Tab | Move cursor to previous tab position |
| LeftShift+Tab | Insert spaces to next tab position |
| RightShift+Tab | Delete blank spaces in the current line “up to and from the cursor” (the manual does not define a more precise selection boundary) |
3. Linear operations on numeric fields
The addendum explicitly limits all operations in Chapter 13 to one- and/or two-dimensional fields with floating-point variables (printed p. 13-1). Do not generalize them to integer/string arrays without evidence.
Index base — printed pp. 13-1–13-2
MAT BASE 0andMAT BASE 1set/reset the row/column offset used by the listed matrix operations.- The text says
MAT BASEis sensibly used whenOPTION BASE 0is active;MAT BASE 1sets the matrix-operation offset to 1, andMAT BASE 0resets it to 0. - The stated default is
MAT BASE 1. - The setting affects
MAT READ,MAT PRINT,MAT CPY,MAT XCPY,MAT ADD,MAT SUB, andMAT MUL.
Generation — printed pp. 13-2–13-3
| Command | Source-described effect |
|---|---|
MAT CLR a() |
Set every element of vector/matrix a() to zero; described as equivalent to ARRAYFILL a(),0 |
MAT SET a()=x |
Set every element to scalar x; described as equivalent to ARRAYFILL a(),x |
MAT ONE a() |
Generate a square identity/unity matrix: diagonal elements 1, all others 0 |
Read, print, and input — printed pp. 13-3–13-4
| Command | Source-described effect |
|---|---|
MAT READ a() |
Read a previously dimensioned vector/matrix from DATA rows (printed p. 13-5) |
MAT PRINT [#i,]a()[g,n] |
Print vector/matrix; vectors on one row with comma-separated elements, matrix rows followed by rowfeed; optional output channel and numeric formatting like STR$(x,g,n) (printed pp. 13-5–13-6) |
MAT INPUT #i,a() |
Read vector/matrix from an ASCII file; the format is the reverse of MAT PRINT, and commas/linefeeds may vary as with INPUT # (printed p. 13-6) |
Copy and transpose — printed pp. 13-4–13-15
| Command | Source-described effect |
|---|---|
MAT CPY a([i,j])=b([k,l])[,h,w] |
Copy a selected rectangular region from b to a |
MAT XCPY a([i,j])=b([k,l])[,h,w] |
Copy while transposing source rows/columns; source matrix remains unchanged |
MAT TRANS a()[=b()] |
Copy transposed matrix b to a; for square matrix, MAT TRANS a() transposes in place |
The page-image review establishes the following exact rules (printed pp. 13-8–13-19):
MAT CPY a([i,j])=b([k,l])[,h,w]copies a rectangle beginning at destination offset(i,j)from source offset(k,l);h,wspecify the copied height and width. The shorthandMAT CPY a()=b()copies the complete overlapping index region when the matrices have different orders: only elements whose row and column indices exist in both source and destination are copied (pp. 13-8–13-10, 13-14–13-15).MAT XCPY a([i,j])=b([k,l])[,h,w]follows the same offset/extent model, but transposesbwhile copying it intoa;bremains unchanged. As withMAT CPY, only elements with matching source/destination indices are copied (p. 13-15).- Vector representation is significant.
DIM a(n)creates a row vector;DIM a(n,1)creates a column vector.MAT CPYtreats both row and column vectors syntactically as column vectors, so its vector copy form isMAT CPY a(n,1)=b(m,1)regardless of the actual vector orientation (p. 13-17). MAT XCPYrequires one operand explicitly dimensioned as a row vector and the other as a column vector; it transposes the second vector before copying. The documented supported cases areDIM a(1,n),b(m,1)(row-vectora, column-vectorb) andDIM a(n,1),b(1,m)(column-vectora, row-vectorb). For a correctMAT XCPYform, the manual says to write the destination/source dimensions as column vectors syntactically (pp. 13-17–13-18).- Optional vector extents have special zero rules (p. 13-19): for
MAT CPY, whenw>=1, onlyhis considered;w=0performs no copy. ForMAT XCPY, when transposing a column vectorbinto a row vector,w>=1makes onlyhrelevant andw=0performs no copy. When transposing a row vectorbinto a column vector,h>=1makes onlywrelevant andh=0performs no copy. MAT TRANS a()=b()copies the transpose ofbintoa; the matrices must have compatible reversed dimensions (ifbisn×m,amust bem×n). The example usesDIM a(3,4),b(4,3). For a square matrix,MAT TRANS a()transposes in place (pp. 13-19–13-20).
These rules are source-verified for the 3.5 addendum and should be preserved explicitly in parser/semantic tests; do not reduce them to a generic slice-copy abstraction.
Arithmetic and matrix analysis — printed pp. 13-15–13-30
| Command form | Source-described effect / constraints |
|---|---|
MAT ADD a()=b()+c() |
Element-wise sum into a; operands must have same order |
MAT ADD a(),b() |
Add b element-wise into a in place; same order; original a lost |
MAT ADD a(),x |
Add scalar x to every element of a in place |
MAT SUB a()=b()-c() |
Element-wise difference into a; operands must have same order |
MAT SUB a(),b() |
Subtract b element-wise from a in place; same order |
MAT SUB a(),x |
Subtract scalar x from every element of a in place |
MAT MUL a()=b()*c() |
If b and c are matrices, multiply rows of b by columns of c; b's column count must equal c's row count, and result a has b's row count and c's column count (printed p. 13-20). If b and c are vectors, this form produces their dyadic/outer product (printed p. 13-21). |
MAT MUL x=a()*b() |
For vectors with equal element counts, calculate scalar/dot (inner) product: sum of pairwise element products (printed p. 13-22). |
MAT MUL x=a()*b()*c() |
Scalar result for a row vector a, compatible matrix b, and column vector c; a has as many elements as b has rows, and c as many elements as b has columns (printed p. 13-22). |
MAT MUL a(),x |
Multiply every element of a by scalar x in place |
MAT NORM a(),0 |
Normalize row-wise: each row is scaled so the sum of squares of its elements is 1 (printed pp. 13-22–13-25). |
MAT NORM a(),1 |
Normalize column-wise: each column is scaled so the sum of squares of its elements is 1 (printed pp. 13-22–13-25). |
MAT DET x=a([i,j])[,n] |
Calculate determinant of a square matrix/selected square submatrix (supplied 469-page manual, printed p. 13-26; combined update scan, pp. 13-34–13-36) |
MAT QDET x=a([i,j])[,n] |
Same determinant purpose and submatrix selection as MAT DET; optimized for speed, not accuracy. The manual recommends MAT DET for critical matrices whose determinant is near zero (supplied 469-page manual, printed pp. 13-27–13-28; combined update scan, pp. 13-36–13-37) |
MAT RANG x=a([i,j])[,n] |
Calculate rank of a square matrix; like MAT DET/MAT QDET, accepts row and column offsets, with n selecting an n×n submatrix from that position (supplied 469-page manual, printed p. 13-29; combined update scan, pp. 13-38–13-39) |
MAT INV b()=a() |
Calculate the inverse of square matrix a and write it to destination matrix b; b() must have the same type/dimensions as a() (printed p. 13-30) |
Page-image checks of printed pp. 13-20–13-22 in the supplied 469-page manual establish the matrix multiplication dimensions. The row/column normalization definitions are on printed pp. 13-22–13-25; the previous registry locator of pp. 13-31–13-33 was inconsistent with the primary manual's page labels and has been corrected. The combined update scan uses a different pagination, cross-referenced in the source/topic index. Printed pp. 13-26–13-30 in the supplied 469-page manual confirm MAT DET, MAT QDET, MAT RANG, and the detailed MAT INV b()=a() example; the combined update scan uses pp. 13-34–13-39 for the corresponding later definitions. Operand direction resolved: the summary on printed p. 13-16 lists the reversed form MAT INV a()=b(), but the detailed definition and worked example on p. 13-30 explicitly state that the inverse of source a() is written to destination b(); the German 1990 book independently confirms destination-first MAT INV i()=m() on p. 475. Treat the summary form as a source typo and use MAT INV destination()=source(). Runtime edge cases remain untested. The earlier OCR-only warning for the matrix multiplication examples is superseded by the page-image checks recorded above. Do not use OCR artifacts as normative syntax.
4. Additional commands in version 3.5 — combined-scan Chapter 14, printed pp. 14-1–14-7
Pagination crosswalk: the supplied 469-page GFA_Basic_3.5_Interpreter.pdf presents the 3.5 addendum as Chapter 13. Its printed pp. 13-33–13-36 contain the FACT/VARIAT/COMBIN, OPENW, PCOLOR, DATA-pointer, special-variable, and GURU material summarized below; _DATA/_DATA= are on p. 13-34 and _0…_9 on p. 13-35. The separately catalogued 487-page GFA-Basic 3.0 + 3.5 Upd (Amiga).pdf divides the same addendum material into Chapter 13 (linear matrix operations) and Chapter 14 (further commands), printed pp. 14-1–14-7; the corresponding special-name passages are on pp. 14-5–14-6. References in this section therefore use the combined scan's Chapter 14 pagination; references to the supplied interpreter PDF must use Chapter 13 pagination. These are two pagination witnesses for the same 3.5 feature material, not evidence for importing Atari or 3.0-only rules.
The combined update scan explicitly says that, in addition to Chapter 13, version 3.5 implements further commands/features (Chapter 14). These are part of the same 3.5 interpreter source and must be included in the compatibility inventory.
| Feature / syntax | Source-described behavior |
|---|---|
x=FACT(n) |
Factorial n!; n is an integer expression, result x is an arithmetic expression; defines 0! = 1 (printed pp. 14-1–14-2) |
y=VARIAT(n,k) |
Number of k-permutations of n elements without repetition: n!/(n-k)!; n,k are integer expressions (printed pp. 14-1–14-2) |
z=COMBIN(n,k) |
Number of k-combinations without repetition: n!/((n-k)!*k!); n,k are integer expressions (printed p. 14-2) |
Page-image review of the supplied interpreter scan's printed p. 13-33 confirms a source-level spelling inconsistency in the worked example: the syntax line and definition repeatedly identify FACT, but the example prints x=FAXT(6). Treat FAXT as an apparent typographical error in that example, not as an alias or a fourth combinatorics function. The documented declaration remains FACT; the page's result example is not used as evidence for a FAXT identifier.
| OPENW with negative screen number | Opens window on the current (top) screen (printed p. 14-3) |
| PCOLOR foreground[,background] | Like COLOR, but applies to screen output produced by PRINT; background is optional in the examples (printed p. 14-4) |
| _DATA | Returns/indicates DATA pointer position; zero if the next READ would cause “out of data” |
| _DATA= | Sets the DATA pointer to a previously determined value obtained from _DATA |
| _0 … _9 | Ten special longword variables, like %, not floating-point; cleared only by NEW, not by CLEAR, LOAD, CHAIN, or RUN (printed p. 14-6) |
| GURU x | Enables/disables GURU checking: x=0 disables it; x<>0 enables it. On a GURU Meditation it transfers to ONERROR GOSUB or generates error #109 (printed pp. 14-6–14-7) |
| GURU | Function returning the GURU number (printed pp. 14-6–14-7) |
Chapter 14's opening page enumerates these as the additional 3.5 features (printed p. 14-1). The examples and definitions are on pp. 14-1–14-6. The closing notes (p. 14-7) state that error messages #110–#118 correspond to normal GURUs $2–$A; in compiled 3.5 programs, RESUME NEXT and RESUME LABEL cannot be executed after an ONERROR GOSUB call caused by a GURU. The manual treats error #109 (GURU Meditation) as a serious system condition and warns that GURU checking only works if the system can continue execution.
Page-image review of the combined update scan's printed p. 14-5 confirms the leading underscore in both DATA-pointer operations: _DATA and _DATA=. Ordinary DATA remains the data declaration statement. Printed p. 14-6 confirms _0…_9 are longword special variables and distinguishes GURU x from the GURU function.
5. Implementation and conformance work breakdown
- Add parser/AST coverage for every MAT statement and Chapter 14 function/command above.
- Implement floating-point-only array constraints and dimension checks from the source; do not silently accept integer arrays.
- Implement and test MAT BASE interaction with OPTION BASE and each affected command.
- Add reference tests for generation, DATA/file I/O, region copy, transpose, arithmetic, norms, determinant, quick determinant, rank, and inverse.
- Add tests for DATA pointer query/restore, special-variable persistence, PCOLOR, negative-screen OPENW, and GURU handling. Tests involving OS crash recovery must be isolated and carefully sandboxed.
- Keep editor behavior separate from language/runtime semantics; it may be documented even if the Rust project does not implement the historical editor.
- Mark all 3.0 manual features not explicitly reconfirmed by this 3.5 manual as inherited candidates, not verified 3.5 facts.
6. Registry declaration crosswalk
The following table maps every declaration in the current 3.5 addendum registry slice to its stable API ID and the printed source locator used by the registry. “Language core” and “Amiga profile” are separate registry namespaces; the latter contains platform-specific PCOLOR, the negative-screen OPENW variant, and the two distinct GURU forms.
Language-core declarations (24)
| Declaration | Stable registry ID | Primary 3.5 printed locator |
|---|---|---|
| MAT BASE | gfa.language.matrix.mat-base | Chapter 13, pp. 13-1–13-2 |
| MAT CLR | gfa.language.matrix.mat-clr | Chapter 13, pp. 13-2–13-3 |
| MAT SET | gfa.language.matrix.mat-set | Chapter 13, pp. 13-2–13-3 |
| MAT ONE | gfa.language.matrix.mat-one | Chapter 13, pp. 13-2–13-3 |
| MAT READ | gfa.language.matrix.mat-read | Chapter 13, p. 13-5 |
| MAT PRINT | gfa.language.matrix.mat-print | Chapter 13, pp. 13-5–13-6 |
| MAT INPUT | gfa.language.matrix.mat-input | Chapter 13, p. 13-6 |
| MAT CPY | gfa.language.matrix.mat-cpy | Chapter 13, pp. 13-4–13-19 |
| MAT XCPY | gfa.language.matrix.mat-xcpy | Chapter 13, pp. 13-15–13-19 |
| MAT ADD | gfa.language.matrix.mat-add | Chapter 13, pp. 13-16–13-18 |
| MAT SUB | gfa.language.matrix.mat-sub | Chapter 13, pp. 13-18–13-20 |
| MAT MUL | gfa.language.matrix.mat-mul | Chapter 13, pp. 13-20–13-22 |
| MAT TRANS | gfa.language.matrix.mat-trans | Chapter 13, pp. 13-19–13-20 |
| MAT NORM | gfa.language.matrix.mat-norm | Chapter 13, pp. 13-22–13-25 |
| MAT DET | gfa.language.matrix.mat-det | Chapter 13, p. 13-26 |
| MAT QDET | gfa.language.matrix.mat-qdet | Chapter 13, pp. 13-27–13-28 |
| MAT RANG | gfa.language.matrix.mat-rang | Chapter 13, p. 13-29 |
| MAT INV | gfa.language.matrix.mat-inv | Chapter 13, p. 13-30 (detailed definition/example) |
| FACT | gfa.language.math.fact | Chapter 13, pp. 13-33–13-34 |
| VARIAT | gfa.language.math.variat | Chapter 13, pp. 13-33–13-34 |
| COMBIN | gfa.language.math.combin | Chapter 13, pp. 13-33–13-34 |
| _DATA | gfa.language.data.data-pointer | Chapter 13, p. 13-34 |
| _DATA= | gfa.language.data.set-data-pointer | Chapter 13, p. 13-34 |
| _0…_9 | gfa.language.memory.special-longword-variables | Chapter 13, p. 13-35 |
Amiga-profile declarations (4)
| Declaration | Stable registry ID | Primary 3.5 printed locator |
|---|---|---|
| PCOLOR | gfa.amiga.graphics.pcolor | Chapter 13, p. 13-34 |
| Negative-screen OPENW variant | gfa.amiga.windows.openw-negative-screen | Chapter 13, p. 13-34 |
| GURU x statement | gfa.amiga.system.guru-check | Chapter 13, p. 13-36 |
| GURU function | gfa.amiga.system.guru-number | Chapter 13, p. 13-36 |
The four Amiga-profile entries are not general language-core declarations; in particular, the GURU statement and GURU function remain distinct declarations under one overload group. The MAT NORM locator above follows the primary manual's page labels on pp. 13-22–13-25; the former registry locator of pp. 13-31–13-33 was incorrect.
This crosswalk is an identity and provenance aid, not an assertion that the registry is a complete 3.5 API catalogue. It does not change implementation, Linux support, or conformance status. The separate corpus fixture crosswalk maps 13 page-image-verified research fixtures to declaration IDs; those fixtures are not executable conformance tests and must not be added to registry support.tests.
Source and rights
This file contains concise original summaries and syntax inventories, not a reproduction of the manual. Cite the source as GFA BASIC 3.5 Interpreter, Chapter 13/14, printed addendum page N. The supplied PDF is a research source and is not committed to the repository.