Current issues in the original source code (In progress)

[tomaslaz]

Major:

• [x] Example Excalibur - does not currently work (since R update) Existing codes needs package rhdf5, which worked on previous R version, but that package doesn't yet(?) exist for updated R: package 'rhdf5' is not available (for R version 3.5.1) https://github.com/alan-turing-institute/DetectorChecker/blob/c020b71ebeb3caa07f17ab3da5d3615314c954e7/Original_code/Rcode_phase1/phase1_PixelReading_test.R#L121-L126 could not find function "H5Fopen"

Possible fix: use package h5

• [x] Layout Excalibur - does not pass the consistency check https://github.com/alan-turing-institute/DetectorChecker/blob/6bec0d8c15ea764fa6cfec0a59086e0f4c01ed6e/Original_code/Rcode_phase0/phase0_LayoutExamples_body.R#L202-L205

• [x] nnorient does not work for all modules See issue #6 for investigation

https://github.com/alan-turing-institute/DetectorChecker/blob/eef5f296f5dc26ece0634cc82014ba7a78b950db/Original_code/Rcode_phase1/phase1_CountsAngles_test.R#L350 Error in (function (x, bw = "nrd0", adjust = 1, kernel = c("gaussian", : 'weights' must all be finite Calls: rose -> nnorient -> circdensity -> do.call -> <Anonymous>

https://github.com/alan-turing-institute/DetectorChecker/blob/eef5f296f5dc26ece0634cc82014ba7a78b950db/Original_code/Rcode_phase1/phase1_CountsAngles_test.R#L402

https://github.com/alan-turing-institute/DetectorChecker/blob/eef5f296f5dc26ece0634cc82014ba7a78b950db/Original_code/Rcode_phase1/phase1_CountsAngles_test.R#L420

This error occurs when all the pixels are close to the module edges rather than other defects. By default the "nnorient" function has "correction" argument set to "bord.modif" which applies edge corrections - those pixels which are closer to edges rather than other pixels are not included in the analysis. A possible fix would be to specify a different correction method or not apply any correction at all, for example correction="none".

The problem has been submitted as a github issue.

Minor:

• [x] readTIFF gives a warning https://github.com/alan-turing-institute/DetectorChecker/blob/c020b71ebeb3caa07f17ab3da5d3615314c954e7/Original_code/Rcode_phase1/phase1_PixelReading_test.R#L65 tiff package currently only supports unsigned integer or float sample formats in direct mode, but the image contains signed integer format - it will be treated as unsigned (use as.is=TRUE, native=TRUE or convert=TRUE depending on your intent)

Fix: as.is = TRUE

[WilfridSKendall]

This message is mainly to check I can add minor issues myself.

I've run the Original_code scripts and noted a few other minor points:

• [ ] Masking of objects: package MASS masks object area from package spatstat. .GlobalEnv masks object edges from package spatstat. Maybe there is nothing that can be done about this.
• [ ] Annoying warning messages about no non-missing arguments to min; returning Inf in (for example) phase1_LogRegression_test.R in Rcode_phase1.