java.lang.NullPointerException: Cannot invoke "java.lang.Integer.intValue()" because "ctx.charCount" is null

[Grufoony]

Describe the bug
java.lang.NullPointerException: Cannot invoke "java.lang.Integer.intValue()" because "ctx.charCount" is null

To Reproduce
Transcribe sheet on a pdf.

Expected behavior
Sheet transcribed correctly.

LOG:

TESSDATA_PREFIX: null
Environment:
- Audiveris:    5.4:c664e2bcc
- OS:           Windows 11 10.0
- Architecture: amd64
- Java VM:      OpenJDK 64-Bit Server VM (build 22.0.2+9-70, mixed mode, sharing)
- OCR Engine:   Tesseract OCR, version 5.3.1
Alias patterns: [(IMSLP[0-9]*)-.*]
Audiveris version 5.4

LogUtil. Configuration found jar:file:/C:/Program%20Files/Audiveris/lib/audiveris.jar!/res/logback.xml
LogUtil. Logging to C:\Users\Gregorio\AppData\Roaming\AudiverisLtd\audiveris\log\20250914T151653.log

Classifier loaded XML norms.
Classifier data loaded from default uri jar:file:/C:/Program%20Files/Audiveris/lib/audiveris.jar!/res/basic-classifier.zip
Versions. Poll frequency: Weekly, next poll on: 21-Sep-2025
Installed OCR languages: eng
1 sheet in C:\Users\Gregorio\Desktop\MyFile.pdf
[MyFile] Loaded image #1 2496x3514 from C:\Users\Gregorio\Desktop\MyFile.pdf
[MyFile] Beam  guessed height: 11 -- 0.50 of 22 interline
[MyFile] Beam measured height: 11 -- 0.31 of [6..22] range at 118% of needed quorum
[MyFile] Scale{ interline(20,22,24) line(2,4,6) beam(11)}
[MyFile] Global slope: -0.01129
[MyFile] Retrieved raw line clusters: 10 of sizes [5] with interline(20,22,24)
[MyFile] Systems: #1[1] #2[2, 3, 4, 5, 6, 7, 8, 9, 10]
[MyFile] Indentation detected for system #1
[MyFile] 9 parts along 2 systems
[MyFile] Created scores: [{Score 1}]
[MyFile] Staff#2 no recognized header clef.
[MyFile] Staff#3 no recognized header clef.
[MyFile] Staff#4 no recognized header clef.
[MyFile] Staff#5 no recognized header clef.
[MyFile] Staff#6 no recognized header clef.
[MyFile] Staff#7 no recognized header clef.
[MyFile] Staff#8 no recognized header clef.
[MyFile] Staff#9 no recognized header clef.
[MyFile] Staff#10 no recognized header clef.
[MyFile] stem(4 max:5)
[MyFile] Scale information: HeadSeeds{NOTEHEAD_BLACK[R:0.4]}
[MyFile] Stems free length median value: 78 pixels, 3.5 interlines
[MyFile] Step CUE_BEAMS is skipped because small heads switch is off
[MyFile]    Abnormal font ratio 3.70 TextWord{ ":;" bounds[2329,672,57,53] conf:0.71 base[2329,719]-[2386,719] codes[3a,3b] FontInfo{Verdana_Bold B-33}}
[MyFile]    Abnormal font ratio 2.08 TextWord{ "z," bounds[1320,3102,37,6] conf:0.37 base[1320,3100]-[1357,3100] codes[7a,2c] FontInfo{Verdana_Bold B-26}}
[MyFile]    Abnormal font ratio 85899345.88 TextWord{ "  " bounds[1742,3380,219,20] conf:0.95 base[1742,3397]-[1961,3397] codes[20,20] FontInfo{Verdana_Bold B-25}}
[MyFile] Metronome. Note characters not found in line: TextLine{ "(A medley including: Superealifragilisticexpialidocious 0 Colonel Hathi‘s March Mickey Mouse March)" bounds[140,254,2113,67] conf:0.80 base[140,314]-[2253,288] Metronome}
[MyFile] Error in performing [SCALE, GRID, HEADERS, STEM_SEEDS, BEAMS, LEDGERS, HEADS, STEMS, REDUCTION, CUE_BEAMS, TEXTS, MEASURES, CHORDS, CURVES, SYMBOLS, LINKS, RHYTHMS, PAGE] java.util.concurrent.ExecutionException: java.lang.RuntimeException: java.lang.NullPointerException: Cannot invoke "java.lang.Integer.intValue()" because "ctx.charCount" is null

[hbitteur]

@Grufoony

I'll be away from my PC for the next 3 weeks.
Sorry!

[Grufoony]

@hbitteur No problem!
I'm just reporting this problem... if you'll need more info in the future i'm going to provide you.

[hbitteur]

@Grufoony

I'm back to "work"! :-)

I noticed you are still using version 5.4, perhaps you could use a more recent version.

Anyway, could you post the input image which led to the error you reported, so that I can investigate.

[Grufoony]

Hi, now i'm the one away! I'll answer to you next week, trying also the new version

[Grufoony]

Hi @hbitteur, i tested it in 5.7.1 and the error persists.
I leave the pdf here

2025_09_14 15_11 Office Lens.pdf

TESSDATA_PREFIX: null
Environment:
- Audiveris:    5.7.1:6579168
- OS:           Windows 11 10.0
- Architecture: amd64
- Java VM:      OpenJDK 64-Bit Server VM (build 24.0.2+12, mixed mode)
- OCR Engine:   Tesseract OCR, version 5.5.1
Alias patterns: [(IMSLP[0-9]*)-.*]
Audiveris version 5.7.1

LogUtil. Configuration found jar:file:/C:/Program%20Files/Audiveris/app/audiveris.jar!/res/logback.xml
LogUtil. Logging to C:\Users\Gregorio\AppData\Roaming\AudiverisLtd\audiveris\log\20251019T142816.log

Classifier loaded XML norms.
Classifier data loaded from default uri jar:file:/C:/Program%20Files/Audiveris/app/audiveris.jar!/res/basic-classifier.zip
Versions. Poll frequency: Weekly, next poll on: 26-Oct-2025
Installed OCR languages: eng
[2025_09_14 15_11 Office Lens] Unmapping is not supported.
[2025_09_14 15_11 Office Lens] 1 sheet in C:\Users\Gregorio\Desktop\2025_09_14 15_11 Office Lens.pdf
[2025_09_14 15_11 Office Lens] Loaded image #1 2496x3514 from C:\Users\Gregorio\Desktop\2025_09_14 15_11 Office Lens.pdf
[2025_09_14 15_11 Office Lens] Beam  guessed height: 11 -- 0.50 of 22 interline
[2025_09_14 15_11 Office Lens] Beam measured height: 11 -- 0.31 of [6..22] range at 119% of needed quorum
[2025_09_14 15_11 Office Lens] Scale{ interline(20,22,24) line(2,4,6) beam(11)}
[2025_09_14 15_11 Office Lens] Global slope: -0.01130
[2025_09_14 15_11 Office Lens] Retrieved raw line clusters: 10 of sizes [5] with interline(20,22,24)
[2025_09_14 15_11 Office Lens] Systems: #1[1] #2[2, 3, 4, 5, 6, 7, 8, 9, 10]
[2025_09_14 15_11 Office Lens] Indentation detected for system #1
[2025_09_14 15_11 Office Lens] 9 parts along 2 systems
[2025_09_14 15_11 Office Lens] Created scores: [{Score 1}]
[2025_09_14 15_11 Office Lens] Staff#2 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#3 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#4 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#5 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#6 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#7 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#8 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#9 no recognized header clef.
[2025_09_14 15_11 Office Lens] Staff#10 no recognized header clef.
[2025_09_14 15_11 Office Lens] stem(4 max:5)
[2025_09_14 15_11 Office Lens] Scale information: HeadSeeds{NOTEHEAD_BLACK[R:0.4]}
[2025_09_14 15_11 Office Lens] Stems free length median value: 78 pixels, 3.5 interlines
[2025_09_14 15_11 Office Lens] Step CUE_BEAMS is skipped because small heads switch is off
[2025_09_14 15_11 Office Lens] Metronome. Note characters not found in line: TextLine{ "(A medley including: Supercalifragilisticexpialidocious 0 Colonel Hathi‘s March . Mickey Mouse March)" bounds[140,254,2113,67] conf:0.82 base[140,314]-[2253,288] Metronome}
[2025_09_14 15_11 Office Lens] Error in performing [SCALE, GRID, HEADERS, STEM_SEEDS, BEAMS, LEDGERS, HEADS, STEMS, REDUCTION, CUE_BEAMS, TEXTS, MEASURES, CHORDS, CURVES, SYMBOLS, LINKS, RHYTHMS, PAGE] java.util.concurrent.ExecutionException: java.lang.RuntimeException: java.lang.NullPointerException: Cannot invoke "java.lang.Integer.intValue()" because "ctx.charCount" is null

[hbitteur]

OK, I found and fixed the reason why a NPE was raised. The engine tried to decipher the line "(A medley including: ...)" as a metronome mark.

Now, even with the fix, we can't get anything interesting out of this input image.
Because all staves, except the very first one, are cropped.
Therefore, we lack the beginning and the end of every staff.

My personal advice: working on such a corrupted input is just a waste of time!

[Grufoony]

I can try to scan it with a better scanner, honestly i didn't think it was such a bad sheet to use (apart from characters, notes appeared clear).

I also don't get why the program seems to see notes (it is highlighting them in green) but it is not understanding what they are.

Given that, I will try a better scan and, finally, do it by hand if it still doesn't work.

Thank you for your feedback and patient!

[hbitteur]

@Grufoony
You are right regarding the notes, they have been correctly recognized.

FYI, the color green has nothing to do with the recognition quality, it just happens to be the color assigned to voice number 2 (with color blue assigned to voice number 1).
See the legend that appears in the middle of Audiveris top line:

Your scan quality looks OK, the problem is the image has been cropped on left and right side. And the program needs to have the beginning and the end of every staff to work correctly, and check the content of every measure (otherwise the erroneous measures are displayed with a pink background, as in your case).

So just make sure you scan the whole sheet width.

Also, a pink background will not prevent you to export the score. It's just a warning telling you something seems wrong in the measure rhythm.

[Etsi-o]

Wouldn't the ability to place start and end staffs be an useful feature? I have many parts with this scan issue and it would be great if they could be recognized even so.

[Grufoony]

@hbitteur
Thank you very much for your feedback.
Unfortunaly, the original print I have has cuts at the border, so this is actually the best I can do.

Following the comment of @Etsi-o, would it possible to implement an auto-deduction for staffs? If the notes are recognized correctly, I think it would not be difficult to sum them up to the correct time signature and I think missing borders would not be a problem anymore.

[hbitteur]

Following the comment of @Etsi-o, would it possible to implement an auto-deduction for staffs? If the notes are recognized correctly, I think it would not be difficult to sum them up to the correct time signature and I think missing borders would not be a problem anymore.

To tell you the truth, I tried to develop this feature several years ago. Depending on the content of the current measure, the engine was sometimes able to infer a time signature.

The key word here is: "sometimes". II had to fight with mis-recognized note heads or rests, augmentation dots, flags, tuplets, implicit tuplets, ... The list is long and the result was unmanageable.

So I decided to rely on few solid foundations, either recognized by the engine or corrected by the end user: the staff header (clef, key, time signature). And then flag every measure whose content was not in line with these foundations.
Doing so, the "contract" between the engine and the end user is clear.

In your concrete case, the input image appears to be cropped on its left side, and you have no other scan available. Fine.

So, you, as the end user, simply have to manually insert the missing items.
This is what I did on your input image:

• First, I cropped the whole vertical line which is located on the image left side (because it was taken as a looooong barline, making all the staves 2 through 10 be part of a huge system).
  I also cropped the horizontal line located at the very top of the image. Perhaps not needed, but since I was cropping the image (via XnView software), I decided to clean it up.
• Second, at then of the HEADERS step, I manually inserted bass clefs at the beginning of staves 2 through 10, paying attention not to overlap with the key signatures.

And then, with a few minor fixes, I came up with this final result:

I did not fix the missing items on the right side of the image.
I'll leave this as an exercise.

[hbitteur]

@Grufoony and @Etsi-o

Since these input images have been truncated on their left and right sides, we need some room to somehow extend the staves manually and populate them with the missing symbols.

For instance, in your particular input case, there was very little room to manually insert the bass clefs (and I had extended each staff as far as possible on its left, a matter of just a few millimeters!).

We could add a feature by which the end user could manually add some white margin around the input image (before inserting any missing symbol).

I have also in mind to let the end user either erase some areas (repainted in white) or simply define some black-listed areas not to be processed by the engine.

What do you guys think of these features?