# Example 695 — The form was accessible and the error was a colour `the_form_was_accessible_and_the_error_was_a_colour.eml` - The form passes every automated accessibility check, a hundred and forty-eight of them. How a user is told a field is wrong is computed below. ## EML ```eml # Self-authored for the EML case corpus (no external origin). The form passes # every automated accessibility check, a hundred and forty-eight of them. How a # user is told a field is wrong is computed below. # # The accessibility work was done and it was not box-ticking. Every input has a # programmatically associated label, the tab order follows the visual order, the # landmarks are right, focus is visible and never trapped, the contrast ratios # exceed the requirement rather than meeting it, and the checker runs in the # pipeline so a regression fails the build. # # What the checker verifies is what a machine can decide from the document. That # an error is CONVEYED is a semantic question: the rule says do not use colour # as the only means of conveying information, and deciding whether something is # the only means requires knowing what the something means. # # A failed field gets a red border. There is no text, no icon, and no # announcement. 148 => automated_checks 148 => automated_checks_passed 15 => fields 15 => fields_whose_error_is_colour_only 41000 => sessions_per_day 450 => colour_vision_deficiency_per_myriad 0 => contrast_failures int(sessions_per_day * colour_vision_deficiency_per_myriad / 10000) => sessions_that_cannot_see_the_signal automated_checks - automated_checks_passed => automated_failures "automated checks : " + str(automated_checks) ^0 "passed : " + str(automated_checks_passed) ^0 "failed : " + str(automated_failures) ^0 "contrast failures : " + str(contrast_failures) ^0 "" ^0 "fields : " + str(fields) ^0 "errors signalled by colour only : " + str(fields_whose_error_is_colour_only) ^0 "sessions per day : " + str(sessions_per_day) ^0 "sessions that cannot see it : " + str(sessions_that_cannot_see_the_signal) ^0 "" ^0 # ---- what the checker verified ---- "the automated suite" ^0 " labels programmatically associated : all " + str(fields) ^0 " tab order follows visual order : yes" ^0 " landmarks : correct" ^0 " focus visible, never trapped : yes" ^0 " contrast : exceeds, not merely meets" ^0 " runs in the pipeline : yes, a regression fails the build" ^0 " verdict : ACCESSIBLE" ^0 "" ^0 " this is a real implementation and the pipeline gate" ^0 " keeps it real" ^0 "" ^0 # ---- what a checker can decide ---- "the two kinds of rule" ^0 " is this contrast ratio above 4.5 : a machine can measure" ^0 " is this label associated : a machine can read" ^0 " is colour the ONLY means of conveying this : requires" ^0 " knowing what is being conveyed" ^0 " so the third is : not in the suite," ^0 " and correctly not in it" ^0 "" ^0 " the checker is not failing to test it; a checker cannot" ^0 " hold the premise the rule is about" ^0 "" ^0 int(fields_whose_error_is_colour_only * 10000 / fields) => colour_only_per_myriad "share of fields whose error is colour only : " + str(colour_only_per_myriad) + " per ten thousand" ^0 "" ^0 # ---- what the user experiences ---- # The submit button does nothing, because the form will not submit while a # field is invalid, and nothing says which field or why. "a failed submission" ^0 " what changes visually : one border becomes red" ^0 " text explaining it : none" ^0 " icon : none" ^0 " announcement to assistive technology : none" ^0 " what a user who cannot see the change experiences :" ^0 " a button that does nothing" ^0 "" ^0 # ---- null control ---- # The same form, with the error also written as text beside the field and # announced. 0 => nc_fields_whose_error_is_colour_only automated_checks_passed => nc_automated_checks_passed "null control - the error also written and announced" ^0 " automated checks passed : " + str(nc_automated_checks_passed) + ", unchanged" ^0 " errors signalled by colour only : " + str(nc_fields_whose_error_is_colour_only) ^0 " the suite did not get stricter; the signal stopped" ^0 " having exactly one channel" ^0 "" ^0 # ---- the rule ---- "what a passing accessibility suite guarantees" ^0 " every machine-decidable rule holds : exactly" ^0 " the interface can be used : not addressed; the" ^0 " rules that need a reader to know what something means" ^0 " are the ones a checker cannot hold, and they are not" ^0 " the easy ones" ^0 "" ^0 "an automated suite partitions the guidelines into the ones it" ^0 "can decide and the ones it cannot, and reports only on the" ^0 "first; a green build is a statement about that partition" ^0 "" ^0 "The form passes all " + str(automated_checks) + " automated checks with " + str(automated_failures) + " failures and " + str(contrast_failures) + " contrast" ^0 "problems, on labels, tab order, landmarks, focus and ratios that exceed rather" ^0 "than meet. All " + str(fields) + " fields signal a validation error with a red border alone -" ^0 str(colour_only_per_myriad) + " per ten thousand - which no checker can decide is the only means, and" ^0 "about " + str(sessions_that_cannot_see_the_signal) + " sessions a day meet a button that appears to do nothing." ^0 ``` ## Python (deterministic transpilation) ```python automated_checks = 148 automated_checks_passed = 148 fields = 15 fields_whose_error_is_colour_only = 15 sessions_per_day = 41000 colour_vision_deficiency_per_myriad = 450 contrast_failures = 0 sessions_that_cannot_see_the_signal = int(sessions_per_day * colour_vision_deficiency_per_myriad / 10000) automated_failures = automated_checks - automated_checks_passed print("automated checks : " + str(automated_checks)) print("passed : " + str(automated_checks_passed)) print("failed : " + str(automated_failures)) print("contrast failures : " + str(contrast_failures)) print("") print("fields : " + str(fields)) print("errors signalled by colour only : " + str(fields_whose_error_is_colour_only)) print("sessions per day : " + str(sessions_per_day)) print("sessions that cannot see it : " + str(sessions_that_cannot_see_the_signal)) print("") print("the automated suite") print(" labels programmatically associated : all " + str(fields)) print(" tab order follows visual order : yes") print(" landmarks : correct") print(" focus visible, never trapped : yes") print(" contrast : exceeds, not merely meets") print(" runs in the pipeline : yes, a regression fails the build") print(" verdict : ACCESSIBLE") print("") print(" this is a real implementation and the pipeline gate") print(" keeps it real") print("") print("the two kinds of rule") print(" is this contrast ratio above 4.5 : a machine can measure") print(" is this label associated : a machine can read") print(" is colour the ONLY means of conveying this : requires") print(" knowing what is being conveyed") print(" so the third is : not in the suite,") print(" and correctly not in it") print("") print(" the checker is not failing to test it; a checker cannot") print(" hold the premise the rule is about") print("") colour_only_per_myriad = int(fields_whose_error_is_colour_only * 10000 / fields) print("share of fields whose error is colour only : " + str(colour_only_per_myriad) + " per ten thousand") print("") print("a failed submission") print(" what changes visually : one border becomes red") print(" text explaining it : none") print(" icon : none") print(" announcement to assistive technology : none") print(" what a user who cannot see the change experiences :") print(" a button that does nothing") print("") nc_fields_whose_error_is_colour_only = 0 nc_automated_checks_passed = automated_checks_passed print("null control - the error also written and announced") print(" automated checks passed : " + str(nc_automated_checks_passed) + ", unchanged") print(" errors signalled by colour only : " + str(nc_fields_whose_error_is_colour_only)) print(" the suite did not get stricter; the signal stopped") print(" having exactly one channel") print("") print("what a passing accessibility suite guarantees") print(" every machine-decidable rule holds : exactly") print(" the interface can be used : not addressed; the") print(" rules that need a reader to know what something means") print(" are the ones a checker cannot hold, and they are not") print(" the easy ones") print("") print("an automated suite partitions the guidelines into the ones it") print("can decide and the ones it cannot, and reports only on the") print("first; a green build is a statement about that partition") print("") print("The form passes all " + str(automated_checks) + " automated checks with " + str(automated_failures) + " failures and " + str(contrast_failures) + " contrast") print("problems, on labels, tab order, landmarks, focus and ratios that exceed rather") print("than meet. All " + str(fields) + " fields signal a validation error with a red border alone -") print(str(colour_only_per_myriad) + " per ten thousand - which no checker can decide is the only means, and") print("about " + str(sessions_that_cannot_see_the_signal) + " sessions a day meet a button that appears to do nothing.") ``` ## stdout (executed) ```text automated checks : 148 passed : 148 failed : 0 contrast failures : 0 fields : 15 errors signalled by colour only : 15 sessions per day : 41000 sessions that cannot see it : 1845 the automated suite labels programmatically associated : all 15 tab order follows visual order : yes landmarks : correct focus visible, never trapped : yes contrast : exceeds, not merely meets runs in the pipeline : yes, a regression fails the build verdict : ACCESSIBLE this is a real implementation and the pipeline gate keeps it real the two kinds of rule is this contrast ratio above 4.5 : a machine can measure is this label associated : a machine can read is colour the ONLY means of conveying this : requires knowing what is being conveyed so the third is : not in the suite, and correctly not in it the checker is not failing to test it; a checker cannot hold the premise the rule is about share of fields whose error is colour only : 10000 per ten thousand a failed submission what changes visually : one border becomes red text explaining it : none icon : none announcement to assistive technology : none what a user who cannot see the change experiences : a button that does nothing null control - the error also written and announced automated checks passed : 148, unchanged errors signalled by colour only : 0 the suite did not get stricter; the signal stopped having exactly one channel what a passing accessibility suite guarantees every machine-decidable rule holds : exactly the interface can be used : not addressed; the rules that need a reader to know what something means are the ones a checker cannot hold, and they are not the easy ones an automated suite partitions the guidelines into the ones it can decide and the ones it cannot, and reports only on the first; a green build is a statement about that partition The form passes all 148 automated checks with 0 failures and 0 contrast problems, on labels, tab order, landmarks, focus and ratios that exceed rather than meet. All 15 fields signal a validation error with a red border alone - 10000 per ten thousand - which no checker can decide is the only means, and about 1845 sessions a day meet a button that appears to do nothing. ``` ## Round-trip `ok: true` — round-trip fixpoint reached (python1 == python2) ## Trace event types eml:run:start · eml:assign · eml:output · eml:run:done