Missing support for some operators in simplified AST

[willend]

I've stumbled upon an issue in mcstas-antlr related to the use of "not equal" in WHEN clauses:
(Have no idea if the issue is specific to WHEN or could arise elsewhere)

2025-01-28 11:43:04.424 | ERROR   | mccode_antlr.reader.reader:syntaxError:26 - Syntax error in parsing Instrument /Users/peterwillendrup/Projects/dmsc-instrumentmodels/estia/mcstas-polarizer/simulation/Estia_baseline.instr at 255,17
2025-01-28 11:43:04.424 | INFO    | mccode_antlr.reader.reader:syntaxError:31 - 
2025-01-28 11:43:04.424 | INFO    | mccode_antlr.reader.reader:syntaxError:31 - /* The actual virtual source mask, two L-shaped absorbers (first top-right) */
2025-01-28 11:43:04.424 | INFO    | mccode_antlr.reader.reader:syntaxError:31 - COMPONENT virtual_source_TR = Slit(
2025-01-28 11:43:04.424 | INFO    | mccode_antlr.reader.reader:syntaxError:31 -   xmin = 0.0, xmax = 1.0, ymin = -1.0, ymax = sample_height/2+over_illumination*5)
2025-01-28 11:43:04.424 | INFO    | mccode_antlr.reader.reader:syntaxError:31 -   WHEN (sample != 4) AT (-over_illumination, 0, -0.5*sample_length) RELATIVE arm_virtual_source
2025-01-28 11:43:04.424 | ERROR   | mccode_antlr.reader.reader:syntaxError:32 - ~~~~~~~~~~~~~~~~~^ mismatched input '!=' expecting ')'
2025-01-28 11:43:04.425 | INFO    | mccode_antlr.reader.reader:syntaxError:34 - 
2025-01-28 11:43:04.425 | INFO    | mccode_antlr.reader.reader:syntaxError:34 - // window to cut down to defined size for test setting
2025-01-28 11:43:04.428 | ERROR   | mccode_antlr.reader.reader:syntaxError:26 - Syntax error in parsing Instrument /Users/peterwillendrup/Projects/dmsc-instrumentmodels/estia/mcstas-polarizer/simulation/Estia_baseline.instr at 265,17
2025-01-28 11:43:04.428 | INFO    | mccode_antlr.reader.reader:syntaxError:31 - //   
2025-01-28 11:43:04.428 | INFO    | mccode_antlr.reader.reader:syntaxError:31 - /* The actual virtual source mask, two L-shaped absorbers (second bottom-left) */
2025-01-28 11:43:04.428 | INFO    | mccode_antlr.reader.reader:syntaxError:31 - COMPONENT virtual_source_BL = Slit(
2025-01-28 11:43:04.428 | INFO    | mccode_antlr.reader.reader:syntaxError:31 -   xmin = -1.0, xmax = 0.0, ymin = -sample_height/2-over_illumination*5, ymax = 1.0)
2025-01-28 11:43:04.428 | INFO    | mccode_antlr.reader.reader:syntaxError:31 -   WHEN (sample != 4) AT (over_illumination, 0, 0.5*sample_length) RELATIVE arm_virtual_source
2025-01-28 11:43:04.428 | ERROR   | mccode_antlr.reader.reader:syntaxError:32 - ~~~~~~~~~~~~~~~~~^ mismatched input '!=' expecting ')'

Version info for mcode-antlr, installed via conda-forge:

(base) yada: $ mcstas-antlr -v
mccode_antlr code generator version 0.9.2
 Copyright (c) European Spallation Source ERIC, 2023-2024
Based on McStas/McXtrace version 3
 Copyright (c) DTU Physics and Risoe National Laboratory, 1997-2023
 Additions (c) Institut Laue Langevin, 2003-2019
All rights reserved

Components are (c) their authors, see component headers.
(base) yada: $ mamba search mccode-antlr
Getting repodata from channels...

mantid/noarch                    ??.?MB @ ??.?MB/s 0.3s
conda-forge/osx-arm64                14.1MB @ 29.4MB/s 0.4s
conda-forge/noarch                 18.8MB @ 34.4MB/s 0.5s
nodefaults/osx-arm64                125.0 B @ 523.0 B/s 0.2s
nodefaults/noarch                  ??.?MB @ ??.?MB/s 0.6s
mantid/osx-arm64                  125.0 B @ 47.0 B/s 2.6s
 Name     Version Build     Channel   Subdir   
─────────────────────────────────────────────────────────────
 mccode-antlr 0.9.2  pyh4b9e410_0  conda-forge conda-forge

[g5t]

Thanks for finding this issue; it is ultimately a result of my incomplete abstract syntax tree construction for expressions -- namely that the binary not-equals operator is omitted here: https://github.com/McStasMcXtrace/mccode-antlr/blob/03f8ce61f2f66e7445cf0809922fd0568717f818/src/grammar/McCommon.g4#L39-L67

The quick-and-dirty workaround is to invert the comparison and negate the result, e.g., WHEN (!(sample == 4)); since both binary equals and unary not are implemented.

I will extend the incomplete AST to include !=, but maybe this warrants a reconsideration of this whole approach

1. The error message produced could be more helpful to indicate what the problem is and how to fix it
2. It should be possible to detect this type of error and fall-back to holding the literal context string instead of attempting to build an AST expression. That could make the instrument compile still at the expense of not being able to do intermediate representation introspection.
3. The other missing operators could be implemented, some of which would require careful consideration (and/or a complete AST)

[g5t]

#99 Fixes your specific issue @willend but doesn't address the indeterminately-sized elephant.

Do you have a good idea of which operators are used in

1. setting instrument parameter values (when they're defined in at the top of the .instr file)
2. assigning component-instance parameters (when the component instance is defined in TRACE)
3. or specifying the argument for SPLIT, WHEN, or JUMP?

Most importantly, do you know if anyone makes use of any of the binary-assignment (=, +=, -=, *=, /=, %=, ^=, &=, |=, <<=, >>=) or unary-modification (++, --) operators in these locations?

[willend]

A first, quick response here @g5t - will come back if I suddenly remember anything else...

1. No operators, effectively only param=value items in the list (I guess you mean "instrument parameters" and not vars in DECLARE/INIT?)
2. Hmm. As such I guess "any", but off the top of my head these are the "normal" ones: +,-,*,/, and maybe unary ! for negation... (We should probably survey the instr suite
3. idem 2, we should survey.

Don't recall any use of those in the final list above - but again we should perhaps survey to be safe.
(On the other hand, it seems the output of the nightly test gives exactly the same for column "mcstas" and "mcstas-antlr" - so we "should" be mostly safe I think...)

[g5t]

The specific missing operator != has been added, but this issue still warrants investigation for others mentioned above.