#!/usr/bin/env bash # Stress test for mdqy. ~110 cases, hunting for parse, eval, # mutation, multi-file, encoder, and jq-compat bugs. # # Usage: scripts/stress.sh (build release if missing, run) # scripts/stress.sh --debug (use dev profile) # scripts/stress.sh -v (print every assertion) # # Exits 0 if every test passes, 1 if any fails. # # Known divergences this script asserts as failures (so regressions # resurface immediately): # - `not` lexes as Tok::KwNot and only parses as a unary prefix, # so `5 | not` is a parse error (jq accepts both forms). # - `try EXPR` form unsupported; only `EXPR?` works. # - `length` on a number errors out; jq returns abs(n). # - `ascii_upcase` / `ascii_downcase` only handle ASCII # (intentional per builtins.rs, divergent from jq). set -u ROOT="$(cd "$(dirname "$0")/.." && pwd)" PROFILE=release VERBOSE=0 for arg in "$@"; do case "$arg" in --debug) PROFILE=debug ;; -v|--verbose) VERBOSE=1 ;; -h|--help) sed -n '2,/^$/p' "$0" | sed 's/^# \{0,1\}//'; exit 0 ;; esac done if [[ "$PROFILE" == "release" ]]; then MDQY="$ROOT/target/release/mdqy" BUILD="--release" else MDQY="$ROOT/target/debug/mdqy" BUILD="" fi if [[ ! -x "$MDQY" ]] || [[ "$ROOT/Cargo.toml" -nt "$MDQY" ]]; then echo "building $PROFILE binary..." >&2 ( cd "$ROOT" && cargo build $BUILD --features tty,watch ) >/dev/null 2>&1 || { echo "build failed"; exit 1 } fi TMP=$(mktemp -d /tmp/mdqy-stress.XXXXXX) trap "rm -rf '$TMP'" EXIT cd "$TMP" || exit 1 PASS=0 FAIL=0 declare -a FAILED if [[ -t 1 ]]; then R=$'\033[31m'; G=$'\033[32m'; Y=$'\033[33m'; D=$'\033[90m'; X=$'\033[0m' else R=''; G=''; Y=''; D=''; X='' fi ok() { PASS=$((PASS + 1)) if (( VERBOSE )); then printf '%sok%s %s\n' "$G" "$X" "$1" else printf '%s.%s' "$G" "$X"; fi } ko() { FAIL=$((FAIL + 1)) FAILED+=("$1") if (( VERBOSE )); then printf '%sFAIL%s %s\n' "$R" "$X" "$1" else printf '%s!%s' "$R" "$X"; fi } # `$()` strips trailing newlines. To preserve exact bytes we append a # sentinel `x` past the binary's output and strip it back. Every # helper that compares output captures the full byte stream this way # so `ts_eq` can match a trailing `\n`. # Substring / equality assertions. `seq_eq` strips a single trailing # newline from each side so that "5" matches mdqy's "5\n" output; # identity tests still match because both source and serialised output # either both end in `\n` or neither does. sin() { [[ "$3" == *"$2"* ]] && ok "$1" || ko "$1 :: want substr='$2' got='${3:0:200}'"; } nin() { [[ "$3" != *"$2"* ]] && ok "$1" || ko "$1 :: forbidden substr='$2' present in '${3:0:200}'"; } seq_eq() { local w="${2%$'\n'}" g="${3%$'\n'}" [[ "$g" == "$w" ]] && ok "$1" || ko "$1 :: want='${2:0:200}' got='${3:0:200}'" } # Stdin-fed tests -------------------------------------------------------- ts_in() { local n="$1" w="$2" s="$3" e="$4"; shift 4 local got=$({ printf '%s' "$s" | "$MDQY" --stdin "$@" "$e" 2>&1; printf x; }) sin "$n" "$w" "${got%x}" } ts_nin() { local n="$1" w="$2" s="$3" e="$4"; shift 4 local got=$({ printf '%s' "$s" | "$MDQY" --stdin "$@" "$e" 2>&1; printf x; }) nin "$n" "$w" "${got%x}" } ts_eq() { local n="$1" w="$2" s="$3" e="$4"; shift 4 local got=$({ printf '%s' "$s" | "$MDQY" --stdin "$@" "$e" 2>&1; printf x; }) seq_eq "$n" "$w" "${got%x}" } ts_fail() { local n="$1" s="$2" e="$3"; shift 3 if printf '%s' "$s" | "$MDQY" --stdin "$@" "$e" >/dev/null 2>&1; then ko "$n :: expected failure, got success" else ok "$n"; fi } # Null-input tests ------------------------------------------------------- tn_in() { local n="$1" w="$2" e="$3"; shift 3 local got=$({ "$MDQY" -n "$@" "$e" 2>&1; printf x; }) sin "$n" "$w" "${got%x}" } tn_eq() { local n="$1" w="$2" e="$3"; shift 3 local got=$({ "$MDQY" -n "$@" "$e" 2>&1; printf x; }) seq_eq "$n" "$w" "${got%x}" } tn_fail() { local n="$1" e="$2"; shift 2 if "$MDQY" -n "$@" "$e" >/dev/null 2>&1; then ko "$n :: expected failure, got success" else ok "$n"; fi } section() { if (( VERBOSE )); then printf '\n%s---- %s ----%s\n' "$Y" "$1" "$X" else printf '\n%s%-25s%s ' "$D" "$1" "$X"; fi } # ---- fixtures -------------------------------------------------------------- TINY=$'# Tiny\n\nA paragraph with [a link](http://example.com).\n\n## Second heading\n\n```rust\nfn main() {}\n```\n' DEEP=$'# A\n\n## A.1\n\n### A.1.1\n\nleaf one.\n\n### A.1.2\n\nleaf two.\n\n## A.2\n\nbody.\n\n# B\n\n## B.1\n\nlast.\n' WITH_FM=$'---\ntitle: Hello\ntags:\n - a\n - b\nnumber: 42\n---\n\n# Body\n\nText.\n' WITH_TOML=$'+++\ntitle = "Hello"\ncount = 7\n+++\n\n# Body\n\nText.\n' UNICODE=$'# Café\n\nGreek: αβγ. Han: 中文. Emoji: 🍕🚀.\n' LIST_DOC=$'# Lists\n\n- one\n- two\n - nested\n- three\n' CODE_DOC=$'# Code\n\n```python\nprint("hi")\n```\n\n```rust\nfn main() {}\n```\n\n```\nno-lang\n```\n' LINKS_DOC=$'# Links\n\n[plain](http://a.com)\n\n[titled](http://b.com "Title here")\n\n![alt-text](http://c.com/img.png "img-title")\n' TABLE_DOC=$'# T\n\n| H1 | H2 |\n| --- | --- |\n| a | b |\n| c | d |\n' # ============================================================================ section "A. identity" ts_eq A_id_tiny "$TINY" "$TINY" '.' ts_eq A_id_deep "$DEEP" "$DEEP" '.' ts_eq A_id_unicode "$UNICODE" "$UNICODE" '.' ts_eq A_id_list "$LIST_DOC" "$LIST_DOC" '.' ts_eq A_id_table "$TABLE_DOC" "$TABLE_DOC" '.' ts_eq A_id_empty "" "" '.' # Bug: walk(.) should be a byte-exact roundtrip on a clean tree. ts_eq A_walk_identity_tiny "$TINY" "$TINY" 'walk(.)' --output md ts_eq A_walk_identity_deep "$DEEP" "$DEEP" 'walk(.)' --output md # ============================================================================ section "B. selectors" ts_in B_h_text "Tiny" "$TINY" 'headings | .text' ts_in B_h_text2 "Second heading" "$TINY" 'headings | .text' ts_in B_h1_text "Tiny" "$TINY" 'h1 | .text' ts_in B_h2_text "Second heading" "$TINY" 'h2 | .text' ts_nin B_h2_no_h1 "Tiny" "$TINY" 'h2 | .text' ts_in B_code_lang "rust" "$TINY" 'codeblocks | .lang' ts_in B_code_literal "fn main" "$TINY" 'codeblocks | .literal' ts_in B_links_href "http://example" "$TINY" 'links | .href' ts_in B_anchor_h1 "tiny" "$TINY" 'h1 | .anchor' ts_in B_anchor_h2 "second-heading" "$TINY" 'h2 | .anchor' ts_in B_paragraphs "paragraph with" "$TINY" 'paragraphs | .text' ts_fail B_h7_unknown "$TINY" 'h7 | .text' ts_in B_lang_filter "rust" "$CODE_DOC" 'codeblocks:lang(rust) | .lang' ts_nin B_lang_no_python "python" "$CODE_DOC" 'codeblocks:lang(rust) | .lang' ts_in B_images_href "img.png" "$LINKS_DOC" 'images | .href' # Image alt text not lifted to attr (bug). The `alt` attr is the # bracket text in markdown image syntax. Test is intentionally strict. ts_in B_images_alt "alt-text" "$LINKS_DOC" 'images | .alt' # ============================================================================ section "C. pseudos" ts_in C_first "Tiny" "$TINY" 'headings:first | .text' ts_in C_last "Second heading" "$TINY" 'headings:last | .text' ts_in C_nth0 "Tiny" "$TINY" 'headings:nth(0) | .text' ts_in C_nth1 "Second heading" "$TINY" 'headings:nth(1) | .text' ts_in C_nth_neg1 "Second heading" "$TINY" 'headings:nth(-1) | .text' ts_eq C_nth_far "null" "$TINY" 'headings:nth(99) | .text' ts_eq C_nth_neg_far "null" "$TINY" 'headings:nth(-99) | .text' ts_in C_text_quote "Tiny" "$TINY" 'headings:text("Tiny") | .text' # ============================================================================ section "D. sections / combinator" ts_in D_section_ascii "Second heading" "$TINY" 'section("Second heading") | .text' ts_in D_section_case "Second heading" "$TINY" 'section("SECOND HEADING") | .text' ts_eq D_section_miss "" "$TINY" 'section("Nope") | .text' --raw ts_in D_hash_title "Tiny" "$TINY" '# Tiny | .text' ts_in D_hash_quoted "Second heading" "$TINY" '## "Second heading" | .text' ts_in D_combinator "fn main" "$TINY" '# "Second heading" > codeblocks | .literal' ts_in D_combinator2 "leaf one" "$DEEP" '# A > # "A.1" > # "A.1.1" | .text' # ============================================================================ section "E. field / index / slice" ts_eq E_kind_root "root" "$TINY" '.kind' --raw ts_in E_text_full "Tiny" "$TINY" '.text' ts_in E_text_para "paragraph" "$TINY" '.text' # .[-1] is the last block-level child. The TINY doc emits # (heading h1, paragraph, heading h2, code). Last is code. ts_eq E_index_neg_kind "\"code\"" "$TINY" '.[-1] | .kind' -c ts_eq E_index_far "null" "$TINY" '.[99]' # Slice ts_in E_slice_kinds "heading" "$TINY" '.[0:1] | .[] | .kind' ts_eq E_slice_inverted "[]" "$TINY" '.[5:1]' -c # Field on null is null tn_eq E_field_null "null" '.foo' # .. recurse all ts_in E_recurse "Second heading" "$TINY" '.. | select(.kind == "heading") | .text' # ============================================================================ section "F. predicates" tn_eq F_select_pass "5" '5 | select(. > 3)' tn_eq F_select_drop "" '5 | select(. > 99)' tn_eq F_any_empty "false" '[] | any' tn_eq F_all_empty "true" '[] | all' tn_eq F_any_one_true "true" '[false, true] | any' tn_eq F_all_one_false "false" '[true, false] | all' # `not` builtin: jq accepts `5 | not`. Mdqy lexes `not` as keyword # (Tok::KwNot) so it can only parse as a unary prefix. Pipe form # parse-errors. Surface the divergence. tn_fail F_not_pipe_form_BUG '5 | not' tn_eq F_not_prefix_null "true" 'not null' tn_eq F_not_prefix_zero "false" 'not 0' tn_eq F_not_prefix_false "true" 'not false' # ============================================================================ section "G. numbers / range" tn_eq G_arith "5" '2 + 3' tn_eq G_div "2" '6 / 3' tn_eq G_mod "1" '7 % 2' tn_eq G_neg "-3" '0 - 3' tn_eq G_int_emit "1" '1.0' tn_eq G_float_emit "1.5" '1.5' tn_eq G_huge "1e+300" '1e300' # JSON formatter normalises exponent tn_eq G_range_pos "[0,1,2,3,4]" '[range(5)]' -c tn_eq G_range_neg "[]" '[range(-3)]' -c tn_eq G_range_step "[5,4,3,2,1]" '[range(5; 0; -1)]' -c tn_fail G_range_zero_step 'range(0; 5; 0)' tn_eq G_limit_2 "[0,1]" '[limit(2; range(10))]' -c tn_eq G_limit_0 "[]" '[limit(0; range(10))]' -c tn_eq G_limit_neg "[]" '[limit(-5; range(10))]' -c tn_eq G_nth_2 "2" 'nth(2; range(10))' tn_eq G_nth_far "null" 'nth(100; range(10))' tn_eq G_nth_neg "null" 'nth(-1; range(10))' # ============================================================================ section "H. strings / regex" tn_eq H_concat '"hello world"' '"hello" + " " + "world"' tn_eq H_split '["a","b","c"]' '"a,b,c" | split(",")' -c tn_eq H_join '"a-b-c"' '["a","b","c"] | join("-")' tn_eq H_upcase '"FOO"' '"foo" | ascii_upcase' tn_eq H_downcase '"bar"' '"BAR" | ascii_downcase' # ascii_upcase only touches ASCII. Documents the behaviour. tn_eq H_upcase_partial '"CAFé"' '"café" | ascii_upcase' tn_eq H_test_match "true" '"hello" | test("h.+o")' tn_eq H_test_no "false" '"hello" | test("xyz")' tn_fail H_test_bad '"x" | test("[unclosed")' tn_eq H_sub '"hxllo hello"' '"hello hello" | sub("e"; "x")' tn_eq H_gsub '"hxllo hxllo"' '"hello hello" | gsub("e"; "x")' tn_eq H_starts "true" '"hello" | startswith("hel")' tn_eq H_ends "true" '"hello" | endswith("llo")' tn_eq H_contains "true" '"hello world" | contains("world")' tn_eq H_ltrimstr '"world"' '"prefix-world" | ltrimstr("prefix-")' tn_eq H_rtrimstr '"world"' '"world-suffix" | rtrimstr("-suffix")' tn_eq H_ltrimstr_id '"world"' '"world" | ltrimstr("xyz")' # Length tn_eq H_len_str "5" '"hello" | length' tn_eq H_len_arr "3" '[1,2,3] | length' tn_eq H_len_obj "2" '{a:1,b:2} | length' tn_eq H_len_null "0" 'null | length' # jq divergence: length(n) = abs(n). Mdqy errors. tn_fail H_len_num_BUG '5 | length' # Interpolation tn_eq H_interp_arith '"x=5"' '"x=\(2+3)"' tn_eq H_interp_field '"name=Alice"' '{name:"Alice"} | "name=\(.name)"' tn_eq H_interp_null '"x=null"' '{} | "x=\(.foo)"' # ============================================================================ section "I. encoders" tn_eq I_csv_simple '"1,\"a, b\",true"' '[1, "a, b", true] | @csv' tn_eq I_csv_quote '"\"\"\"hi\"\"\""' '["\"hi\""] | @csv' tn_eq I_tsv '"a\tb\tc"' '["a","b","c"] | @tsv' tn_eq I_uri_space '"hello%20world"' '"hello world" | @uri' tn_eq I_uri_unicode '"caf%C3%A9"' '"café" | @uri' tn_eq I_html_lt '"<tag>"' '"" | @html' tn_eq I_html_amp '"a&b"' '"a&b" | @html' tn_eq I_html_quote '"'hi'"' $'"\047hi\047" | @html' tn_in I_sh_works "hello" '"hello" | @sh' tn_in I_sh_apostrophe "\\'" '"don'\''t" | @sh' tn_eq I_json_compact '"[1,2,3]"' '[1,2,3] | @json' # ============================================================================ section "J. object / array" tn_eq J_obj '{"a":1,"b":2}' '{a:1, b:2}' -c tn_eq J_obj_short '{"x":5}' '{x:5} | {x}' -c tn_eq J_obj_compkey '{"NAME":"v"}' '{("NAME"): "v"}' -c tn_eq J_arr '[1,2,3]' '[1,2,3]' -c tn_eq J_keys_obj '["a","b"]' '{b:2,a:1} | keys' -c tn_eq J_keys_arr '[0,1,2]' '[10,20,30] | keys' -c tn_eq J_has_obj_yes "true" '{x:1} | has("x")' tn_eq J_has_obj_no "false" '{x:1} | has("y")' tn_eq J_has_arr_yes "true" '[1,2,3] | has(1)' tn_eq J_has_arr_no "false" '[1,2,3] | has(99)' tn_eq J_add_nums "6" '[1,2,3] | add' tn_eq J_add_strs '"abc"' '["a","b","c"] | add' tn_eq J_add_arrs '[1,2,3,4]' '[[1,2],[3,4]] | add' -c tn_eq J_add_empty "null" '[] | add' tn_eq J_sort '[1,2,3]' '[3,1,2] | sort' -c tn_eq J_sort_mixed '[null,false,1,"a"]' '[1, "a", null, false] | sort' -c tn_eq J_sort_by_len '["a","bb","ccc"]' '["bb","a","ccc"] | sort_by(length)' -c tn_eq J_unique '[1,2,3]' '[3,1,2,1,3] | unique' -c tn_eq J_group_by '[[2],[1,3]]' '[1,2,3] | group_by(. % 2)' -c tn_eq J_min_by '"a"' '["bb","a","ccc"] | min_by(length)' tn_eq J_max_by '"ccc"' '["bb","a","ccc"] | max_by(length)' tn_eq J_min "1" '[3,1,2] | min' tn_eq J_max "3" '[3,1,2] | max' tn_eq J_min_empty "null" '[] | min' tn_eq J_slice_neg '[2,3]' '[1,2,3] | .[-2:]' -c tn_eq J_slice_pos '[1]' '[1,2,3] | .[:1]' -c tn_eq J_slice_oob '[1,2,3]' '[1,2,3] | .[0:99]' -c tn_eq J_slice_inv '[]' '[1,2,3] | .[2:1]' -c # ============================================================================ section "K. paths" tn_eq K_paths_obj '[["a"]]' '{a:1} | [paths]' -c tn_eq K_paths_nested '[["a"],["a","b"]]' '{a:{b:1}} | [paths]' -c tn_eq K_getpath "1" '{a:{b:1}} | getpath(["a","b"])' tn_eq K_getpath_miss "null" '{a:1} | getpath(["x","y"])' tn_eq K_getpath_empty '{"a":1}' '{a:1} | getpath([])' -c tn_eq K_setpath_new '{"a":{"b":5}}' 'null | setpath(["a","b"]; 5)' -c tn_eq K_setpath_over '{"a":{"b":2}}' '{a:{b:1}} | setpath(["a","b"]; 2)' -c tn_eq K_setpath_arr '[null,null,"x"]' '[] | setpath([2]; "x")' -c tn_eq K_setpath_empty "5" '{} | setpath([]; 5)' # ============================================================================ section "L. type / conv" tn_eq L_type_null '"null"' 'null | type' tn_eq L_type_str '"string"' '"x" | type' tn_eq L_type_num '"number"' '5 | type' tn_eq L_type_bool '"boolean"' 'true | type' tn_eq L_type_arr '"array"' '[] | type' tn_eq L_type_obj '"object"' '{} | type' tn_eq L_tos_null '"null"' 'null | tostring' tn_eq L_tos_int '"5"' '5 | tostring' tn_eq L_tos_float '"1.5"' '1.5 | tostring' tn_eq L_tos_bool '"true"' 'true | tostring' tn_eq L_tos_arr '"[1,2,3]"' '[1,2,3] | tostring' tn_eq L_tos_obj '"{\"a\":1}"' '{a:1} | tostring' tn_eq L_tos_str '"keep"' '"keep" | tostring' tn_eq L_ton_str "42" '"42" | tonumber' tn_eq L_ton_float "3.14" '"3.14" | tonumber' tn_fail L_ton_bad '"abc" | tonumber' tn_fail L_ton_bool 'true | tonumber' tn_eq L_toj_obj '"{\"a\":1}"' '{a:1} | tojson' tn_eq L_fromj '{"a":1}' '"{\"a\":1}" | fromjson' -c tn_fail L_fromj_bad '"{not json" | fromjson' # ============================================================================ section "M. mutation" DOC_HTTP=$'See [docs](http://example.com).\n' DOC_HTTPS=$'See [docs](https://example.com).\n' ts_eq M_link_rewrite "$DOC_HTTPS" "$DOC_HTTP" \ '(.. | select(type == "link")).href |= sub("http:"; "https:")' --output md ts_eq M_link_idempotent "$DOC_HTTPS" "$DOC_HTTPS" \ '(.. | select(type == "link")).href |= sub("http:"; "https:")' --output md DOC_TITLED=$'See [docs](https://example.com "title").\n' ts_nin M_del_title "title" "$DOC_TITLED" \ 'del((.. | select(type == "link")).title)' --output md # Mutation with no targets: identity output. ts_eq M_no_targets "$TINY" "$TINY" \ '(.. | select(type == "image")).href |= sub("http:"; "https:")' --output md # Walk that bumps levels. ts_in M_walk_bump "## Tiny" "$TINY" \ 'walk(if .kind == "heading" then .level |= . + 1 else . end)' --output md # walk(.) keeps text content (even if not byte-exact). ts_in M_walk_text "Tiny" "$TINY" 'walk(.)' --output md # del then output: title gone from emitted markdown. ts_nin M_del_title_in_output "title" "$DOC_TITLED" \ 'del((.. | select(type == "link")).title)' --output md # `=` (set) still NotImplemented; should fail when run via -U. TFI=$(mktemp "$TMP/setfail.XXXXXX.md"); printf '%s' "$DOC_HTTP" > "$TFI" "$MDQY" -U '.text = "x"' "$TFI" >/dev/null 2>&1 \ && ko "M_set_assign_fails :: -U with `=` should error" \ || ok M_set_assign_fails # Unknown attr name on a mutation target: fail. TFA=$(mktemp "$TMP/badattr.XXXXXX.md"); printf '%s' "$TINY" > "$TFA" "$MDQY" -U 'h1.bogus_attr |= "x"' "$TFA" >/dev/null 2>&1 \ && ko "M_unknown_attr_fails :: bogus attr should error" \ || ok M_unknown_attr_fails # Mutation on --stdin should not silently no-op (potential bug). got_mut=$(printf '%s' "$DOC_HTTP" | "$MDQY" --stdin --output md \ '(.. | select(type == "link")).href |= sub("http:"; "https:")' 2>&1) if [[ "$got_mut" == *"https://"* ]]; then ok M_stdin_mutation_runs else ko "M_stdin_mutation_runs :: stdin+mutation produced '${got_mut:0:80}'" fi # ============================================================================ section "N. multi-file & flags" mkdir -p docs printf '# A\n\nA body.\n' > docs/a.md printf '# B\n\nB body.\n' > docs/b.md printf '# Hidden\n' > docs/.hidden.md echo 'docs/.hidden.md' > .ignore got=$("$MDQY" 'headings | .text' docs/ 2>&1) [[ "$got" == *"A"* && "$got" == *"B"* ]] && ok N_per_file || ko "N_per_file :: '$got'" got=$("$MDQY" --slurp 'length' docs/ 2>&1) [[ "$got" == "2" ]] && ok N_slurp || ko "N_slurp :: '$got'" got=$("$MDQY" --merge 'headings | .text' docs/ 2>&1) [[ "$got" == *"A"* && "$got" == *"B"* ]] && ok N_merge || ko "N_merge :: '$got'" got=$("$MDQY" --no-ignore --hidden --slurp 'length' docs/ 2>&1) [[ "$got" == "3" ]] && ok N_no_ignore_hidden || ko "N_no_ignore_hidden :: '$got'" got_s=$("$MDQY" --workers 1 'headings | .text' docs/ 2>&1) got_p=$("$MDQY" --workers 4 'headings | .text' docs/ 2>&1) [[ "$got_s" == "$got_p" ]] && ok N_workers_match || ko "N_workers_match :: serial='$got_s' parallel='$got_p'" got=$("$MDQY" -n '1+2'); [[ "$got" == "3" ]] && ok N_null_input || ko "N_null_input :: '$got'" got=$(printf '%s' "$TINY" | "$MDQY" --stdin --raw '.text' 2>&1) [[ "$got" != *'"'* ]] && ok N_raw_strips_quotes || ko "N_raw_strips_quotes :: '$got'" got=$(printf 'hello world' | "$MDQY" -R --stdin '.') [[ "$got" == *"hello world"* ]] && ok N_raw_input || ko "N_raw_input :: '$got'" got=$("$MDQY" -n --arg name "Bob" '"hi \($name)"') [[ "$got" == *"hi Bob"* ]] && ok N_arg || ko "N_arg :: '$got'" got=$("$MDQY" -n --argjson n 7 '$n + 3') [[ "$got" == "10" ]] && ok N_argjson || ko "N_argjson :: '$got'" "$MDQY" -n --argjson n 'not-json' '$n' >/dev/null 2>&1 \ && ko "N_argjson_bad :: should fail" || ok N_argjson_bad "$MDQY" --compile-only 'not | valid |' >/dev/null 2>&1 \ && ko "N_compile_bad :: should fail" || ok N_compile_bad got=$("$MDQY" --explain-mode 'headings | .text') [[ "$got" == "mode: stream" ]] && ok N_explain_stream || ko "N_explain_stream :: '$got'" got=$("$MDQY" --explain-mode '[headings] | length') [[ "$got" == "mode: tree" ]] && ok N_explain_tree || ko "N_explain_tree :: '$got'" "$MDQY" --slurp --merge '.' docs/ >/dev/null 2>&1 \ && ko "N_slurp_merge_conflict :: should fail" || ok N_slurp_merge_conflict got=$(printf '%s' "$TINY" | "$MDQY" --output text --stdin '.') [[ "$got" == *"Tiny"* && "$got" != *"#"* ]] && ok N_output_text \ || ko "N_output_text :: '${got:0:80}'" TF=$(mktemp "$TMP/inplace.XXXXXX.md") printf 'See [x](http://e.com).\n' > "$TF" "$MDQY" -U '(.. | select(type == "link")).href |= sub("http:"; "https:")' "$TF" 2>&1 >/dev/null [[ "$(cat "$TF")" == *"https://"* ]] && ok N_in_place || ko "N_in_place :: '$(cat "$TF")'" TB=$(mktemp "$TMP/backup.XXXXXX.md") printf 'See [x](http://e.com).\n' > "$TB" "$MDQY" -U --backup bak '(.. | select(type == "link")).href |= sub("http:"; "https:")' "$TB" 2>&1 >/dev/null if compgen -G "${TB}.bak" >/dev/null || compgen -G "${TB%.md}.md.bak" >/dev/null; then ok N_backup_made else ko "N_backup_made :: no backup near $TB ($(ls -1 "$TMP" | grep -i backup))" fi # ============================================================================ section "O. compile errors" ts_fail O_paren '' '(. ' ts_fail O_brack '' '[. ' ts_fail O_brace '' '{ a: ' ts_fail O_trailing '' '. |' ts_fail O_double_pipe '' '. || .' ts_fail O_unterm_str '' '"oops' ts_fail O_pseudo "$TINY" 'headings:bogus | .text' ts_fail O_unknown_fn "$TINY" 'thiss_does_not_exist' # `try EXPR` form not supported (only `EXPR?`). Mark as bug. tn_fail O_try_form_BUG '' 'try error("bang")' # But `?` postfix works. tn_eq O_try_postfix "" 'error("bang")?' # ============================================================================ section "P. frontmatter" ts_in p_yaml_title "Hello" "$WITH_FM" 'frontmatter | .title' ts_in p_yaml_tag "a" "$WITH_FM" 'frontmatter | .tags | .[0]' --raw ts_in p_yaml_num "42" "$WITH_FM" 'frontmatter | .number' ts_in p_toml_title "Hello" "$WITH_TOML" 'frontmatter | .title' ts_in p_toml_count "7" "$WITH_TOML" 'frontmatter | .count' ts_eq p_fm_missing "null" "$TINY" 'frontmatter' # ============================================================================ section "Q. stream/tree parity" # For each stream-eligible expression, run it through the binary # (which dispatches to stream mode) and through the binary in tree # mode (forced by wrapping in a noop that has_mutation rejects). # Outputs must match exactly. parity() { local name="$1" doc="$2" expr="$3" local out_a out_b out_a=$({ printf '%s' "$doc" | "$MDQY" --stdin "$expr" 2>&1; printf x; }) out_b=$({ printf '%s' "$doc" | "$MDQY" --stdin "[$expr] | .[]" 2>&1; printf x; }) if [[ "${out_a%x}" == "${out_b%x}" ]]; then ok "$name" else ko "$name :: stream='${out_a%x}' tree='${out_b%x}'" fi } parity Q_h_text "$DEEP" 'headings | .text' parity Q_h_anchor "$DEEP" 'headings | .anchor' parity Q_h_lvl_filter "$DEEP" 'headings | select(.level == 2) | .text' parity Q_code_lang "$CODE_DOC" 'codeblocks | .lang' parity Q_code_lit "$CODE_DOC" 'codeblocks | .literal' parity Q_links_href "$LINKS_DOC" 'links | .href' # ============================================================================ section "R. JSON schema" ts_in R_json_kind '"kind":"heading"' "$TINY" '.. | select(.kind == "heading")' --output json -c ts_in R_json_text '"text":"Tiny"' "$TINY" '.. | select(.kind == "heading")' --output json -c ts_in R_json_int_lvl '"level":1' "$TINY" '.. | select(.kind == "heading")' --output json -c ts_nin R_json_no_float '1.0' "$TINY" '.. | select(.kind == "heading")' --output json -c ts_nin R_json_no_span '"span":' "$TINY" '.. | select(.kind == "heading")' --output json -c ts_in R_json_with_span '"span":' "$TINY" '.. | select(.kind == "heading")' --output json --with-spans -c ts_nin R_json_no_empty_children '"children":[]' "$TINY" 'h1' --output json -c # ============================================================================ section "S. edges" ts_eq S_empty_id "" "" '.' ts_eq S_empty_kind "root" "" '.kind' --raw ts_eq S_empty_text "" "" '.text' --raw DOC_FM_ONLY=$'---\nx: 1\n---\n' ts_in S_fm_only_x "1" "$DOC_FM_ONLY" 'frontmatter | .x' DOC_RICH=$'# Hello **bold** `code`\n\nbody.\n' ts_in S_rich "Hello bold code" "$DOC_RICH" 'h1 | .text' DOC_SETEXT=$'Setext\n======\n\npara\n' ts_in S_setext "Setext" "$DOC_SETEXT" 'h1 | .text' # Try / error tn_eq S_try_postfix "" 'error("bang")?' tn_fail S_no_try_fails 'error("bang")' # Reduce / foreach tn_eq S_reduce_sum "10" '[1,2,3,4] | reduce .[] as $x (0; . + $x)' tn_eq S_foreach "[1,3,6,10]" '[1,2,3,4] | [foreach .[] as $x (0; . + $x; .)]' -c # User defs tn_eq S_def "12" 'def double(x): x + x; double(6)' tn_eq S_def_two "12" 'def add2(x; y): x + y; add2(5; 7)' # As binding tn_eq S_as "10" '5 as $x | $x + $x' # Undefined variable tn_fail S_undef '$undefined' # Recurse on nested # `..` walks every value including the top. tn_eq S_recurse_full "[[[1,2]],[1,2],1,2]" '[[1,2]] | [..]' -c # Comma operator tn_eq S_comma '[1,2,3]' '[1,2,3]' -c # Trailing newline preserved through identity DOC_NL=$'# x\n' ts_eq S_trailing_nl "$DOC_NL" "$DOC_NL" '.' # CRLF input survives identity DOC_CRLF=$'# x\r\n\r\nbody\r\n' ts_eq S_crlf "$DOC_CRLF" "$DOC_CRLF" '.' # ============================================================================ section "T. pathological markdown" # Deeply nested blockquote, identity must round-trip byte-exact. DOC_QQ=$'> > > > > > > deep\n' ts_eq T_qq_id "$DOC_QQ" "$DOC_QQ" '.' # 10-level nested list. Indentation matters; if pulldown-cmark # reshapes it, the round-trip fails and surfaces the bug. DOC_NESTED_LIST=$'- a\n - b\n - c\n - d\n - e\n - f\n - g\n - h\n - i\n - j\n' ts_eq T_list_10_id "$DOC_NESTED_LIST" "$DOC_NESTED_LIST" '.' # Aligned table columns. Identity round-trip. DOC_ALIGN=$'| L | C | R |\n| :--- | :---: | ---: |\n| a | b | c |\n' ts_eq T_align_id "$DOC_ALIGN" "$DOC_ALIGN" '.' ts_in T_align_table_kind "table" "$DOC_ALIGN" 'tables | .kind' # Malformed table: missing separator row. Should still parse as paragraph. DOC_BAD_TABLE=$'| a | b |\n| c | d |\n' ts_in T_bad_table "paragraph" "$DOC_BAD_TABLE" '.[0] | .kind' --raw # Reference-style links + footnotes. DOC_REF=$'See [docs][r] and a footnote[^1].\n\n[r]: http://r.com\n[^1]: a note\n' ts_in T_ref_link "http://r.com" "$DOC_REF" 'links | .href' ts_in T_footnote "a note" "$DOC_REF" 'footnotes | .text' # Definition list (GFM extension is on). DOC_DL=$'Term\n: Definition\n' ts_in T_deflist "Term" "$DOC_DL" '.. | select(.kind == "list") | .text' # Hard break in heading: trailing two spaces. Pulldown-cmark may # refuse hard breaks inside headings; identity should still survive. # Identity test only, since rules vary by parser version. DOC_HD_HEADING=$'# Title \nstuff\n' ts_eq T_hd_heading_id "$DOC_HD_HEADING" "$DOC_HD_HEADING" '.' # Tab-indented code (4-space-equivalent indent block). DOC_TAB_CODE=$'\tindented code\n' ts_eq T_tab_code_id "$DOC_TAB_CODE" "$DOC_TAB_CODE" '.' # Mixed tab + space indent. Identity test. DOC_MIX_INDENT=$'-\ta\n- \tb\n' ts_eq T_mix_id "$DOC_MIX_INDENT" "$DOC_MIX_INDENT" '.' # CR-only line endings. Pulldown-cmark may not recognise old-Mac newlines. DOC_CR=$'# x\rbody\r' ts_eq T_cr_id "$DOC_CR" "$DOC_CR" '.' # No trailing newline at all. DOC_NO_NL=$'# x\n\nno trailing newline' ts_eq T_no_nl_id "$DOC_NO_NL" "$DOC_NO_NL" '.' # Trailing whitespace at line end (markdown hard break trigger). DOC_TRAILING_WS=$'line one \nline two\n' ts_eq T_trailing_ws_id "$DOC_TRAILING_WS" "$DOC_TRAILING_WS" '.' # BOM at file start. `--stdin` reads raw UTF-8 — does mdqy strip # the BOM or pass it through? DOC_BOM=$'\xef\xbb\xbf# Title\n\nbody.\n' ts_eq T_bom_id "$DOC_BOM" "$DOC_BOM" '.' # Wide table with many columns. Round-trip should hold. DOC_WIDE=$'| a | b | c | d | e | f | g |\n| - | - | - | - | - | - | - |\n| 1 | 2 | 3 | 4 | 5 | 6 | 7 |\n' ts_eq T_wide_id "$DOC_WIDE" "$DOC_WIDE" '.' # ============================================================================ section "U. extensions" # GFM strikethrough: `~~x~~` should produce a strikethrough kind. DOC_STRIKE=$'~~gone~~\n' ts_in U_strike_id "$DOC_STRIKE" "$DOC_STRIKE" '.' ts_in U_strike_kind "strikethrough" "$DOC_STRIKE" '.. | .kind' # Task lists. DOC_TASKS=$'- [x] done\n- [ ] open\n' ts_eq U_tasks_id "$DOC_TASKS" "$DOC_TASKS" '.' # Test if the .checked attr is exposed via JSON output. ts_in U_tasks_check_true "true" "$DOC_TASKS" '.. | select(.kind == "item") | .checked' -c # Smart punctuation: `--` should become en/em dash on render. DOC_SMART=$'Hello -- world.\n' ts_in U_smart_id "Hello" "$DOC_SMART" '.text' # Wikilinks (GFM extension on). DOC_WIKI=$'See [[Page Name]] for more.\n' ts_in U_wiki_id "Page Name" "$DOC_WIKI" '.text' # Math: inline `$...$`. Result should round-trip (events::options ON). DOC_MATH=$'Equation: $a^2 + b^2 = c^2$.\n' ts_in U_math_id "a^2" "$DOC_MATH" '.text' # Display math: `$$...$$`. DOC_DMATH=$'$$a + b$$\n' ts_in U_dmath_id "a + b" "$DOC_DMATH" '.text' # Heading attributes `# Title {#anchor}`. Pulldown-cmark exposes `id` # but mdqy reads it into `attr::ANCHOR` only if the parser surfaces it. DOC_HID=$'# Welcome {#welcome-id}\n' ts_in U_anchor_attr "welcome-id" "$DOC_HID" 'h1 | .anchor' # ============================================================================ section "V. frontmatter edges" # Empty frontmatter `---\n---`. Should parse without crashing. DOC_FM_EMPTY=$'---\n---\n# Body\n' ts_in V_fm_empty_body "Body" "$DOC_FM_EMPTY" 'h1 | .text' # Frontmatter must be at file start; one in middle should be ignored # but pulldown-cmark + ENABLE_*_METADATA_BLOCKS still parses it. DOC_FM_MID=$'# Body\n\n---\nx: 1\n---\n' ts_eq V_fm_mid_null_BUG "null" "$DOC_FM_MID" 'frontmatter' # Malformed YAML body — frontmatter attr stays unset, returns null. DOC_FM_BAD=$'---\nfoo: : :\n---\n# B\n' ts_eq V_fm_bad_null "null" "$DOC_FM_BAD" 'frontmatter' # TOML with arrays + tables. DOC_TOML_RICH=$'+++\ntags = ["a", "b"]\n[author]\nname = "Bob"\n+++\n# B\n' ts_in V_toml_array "a" "$DOC_TOML_RICH" 'frontmatter | .tags | .[0]' --raw ts_in V_toml_table "Bob" "$DOC_TOML_RICH" 'frontmatter | .author | .name' # Frontmatter only, no body. Identity round-trip. DOC_FM_NO_BODY=$'---\nx: 1\n---\n' ts_eq V_fm_no_body_id "$DOC_FM_NO_BODY" "$DOC_FM_NO_BODY" '.' # `---` inside a YAML string. The closing fence is on its own line so # pulldown-cmark stops at the right place. DOC_FM_INNER=$'---\ntext: "a --- b"\n---\n# B\n' ts_in V_fm_inner_dashes "a --- b" "$DOC_FM_INNER" 'frontmatter | .text' # ============================================================================ section "W. lex / parse edges" # Long identifier (1000 chars). Should compile a no-such-builtin error # rather than a lex one. LONGID=$(printf 'a%.0s' {1..1000}) ts_fail W_long_ident "" "$LONGID" # All supported escapes in a string. tn_eq W_esc_quote '"\""' '"\""' tn_eq W_esc_back '"\\"' '"\\"' tn_eq W_esc_slash '"/"' '"\/"' tn_eq W_esc_n '"\n"' '"\n"' tn_eq W_esc_t '"\t"' '"\t"' tn_eq W_esc_r '"\r"' '"\r"' # `\0` lexes to NUL; JSON formatter emits ``. tn_in W_esc_zero_unicode "u0000" '"\0" | tojson' # '""' '"\0"' # `\u` escape is NOT in the accepted escape set; lex must reject it. # We construct the string literally so Edit-tool quoting can't elide # the backslash-u sequence. ESC_U_EXPR=$(printf '"%s%s%s"' '\' 'u0041') tn_fail W_esc_u_unsupported "$ESC_U_EXPR" # Unicode in identifiers should NOT be allowed (ASCII only). tn_fail W_unicode_ident 'café' # Integer overflow yields infinity per IEEE-754. JSON emits null. tn_eq W_overflow_inf "null" '1e308 * 1e308' # Negative zero compares equal to zero. tn_eq W_neg_zero "true" '(-0) == 0' # Scientific notation lexes. tn_eq W_sci_exp "150" '1.5e2' # 1000 nested parens. Lex/parse should handle without stack-overflow. LP=$(printf '(%.0s' {1..1000}); RP=$(printf ')%.0s' {1..1000}) tn_eq W_deep_paren "1" "${LP}1${RP}" # Alt operator chaining: right-fold value semantics. `null // null // 5`. tn_eq W_alt_chain "5" 'null // null // 5' # `if/elif/else/end` chain. tn_eq W_if_elif '"two"' 'if 1 == 0 then "one" elif 2 == 2 then "two" else "n" end' # def with overload by arity. Mdqy: same name with different arity. # May or may not resolve; pin the behaviour. tn_eq W_def_arity "11" 'def f(g): g + 1; def f(g; h): g + h; f(5; 6)' # Nested defs. mdqy parses `def name:` as a top-level form only — a # nested `def` mid-pipeline isn't accepted by parse_top. tn_eq W_nested_def_BUG "12" 'def outer(x): def inner(y): y * 2; inner(x); outer(6)' # Recursive def: the body refers to itself. mdqy currently # stack-overflows trying to evaluate it. tn_eq W_recursive_def_BUG "120" 'def fact(n): if n < 2 then 1 else n * fact(n-1) end; fact(5)' # Shadowed `as` binding. tn_eq W_shadow_as "7" '5 as $x | 7 as $x | $x' # ============================================================================ section "X. interpolation" # Two interpolations. tn_eq X_two_interp '"x=1 y=2"' '{x:1,y:2} | "x=\(.x) y=\(.y)"' # Nested interpolation: outer string contains inner string with # interp. parse.rs::find_matching_paren walks string literals as a # unit, but lex_string then re-runs over the inner expr and chokes # on the lone backslash before `(`. tn_eq X_nested_interp_BUG '"hi=A"' '"A" as $a | "hi=\("\($a)")"' # Interp containing a pipe. tn_eq X_interp_pipe '"len=5"' '"hello" | "len=\(. | length)"' # Interp inside pseudo-arg `:text(...)`. The pseudo's parse_pipeline # eats the literal as-is, so the `\(.foo)` is taken as the expected # string; no heading text matches. ts_eq X_interp_pseudo_no_subst "" "$TINY" 'headings:text("\(.foo)") | .text' --raw # Empty interp `\()` — find_matching_paren returns immediately. # Should be a parse error. tn_fail X_empty_interp_BUG '"\()"' # Interp followed by literal text. tn_eq X_interp_then_lit '"x=1!"' '{x:1} | "x=\(.x)!"' # ============================================================================ section "Y. mutation depth" # Walk that bumps only h1 levels, leaves h2 alone. DOC_HH=$'# A\n\n## B\n' DOC_HH_BUMP=$'## A\n\n## B\n' ts_eq Y_walk_cond_bump "$DOC_HH_BUMP" "$DOC_HH" \ 'walk(if .kind == "heading" and .level == 1 then .level |= . + 1 else . end)' --output md # Nested walk: walk(walk(.)). The outer mutate.rs handles `walk`, # but the inner `walk` is dispatched through eval where `walk` # isn't registered — runtime error. ts_eq Y_walk_walk_id_BUG "$TINY" "$TINY" 'walk(walk(.))' --output md # Mutation inside `if` then-branch (no else). ts_in Y_if_no_else "https" "$DOC_HTTP" \ 'walk(if .kind == "link" then .href |= sub("http:"; "https:") else . end)' --output md # Mutation that changes attr to wrong type. `.level |= "x"` puts a # string into a slot that the serializer expects to be a number. # mdqy silently keeps the original output instead of erroring or # producing junk — wrong-type writes go unnoticed. TFW=$(mktemp "$TMP/wrongtype.XXXXXX.md"); printf '# Title\n' > "$TFW" out=$("$MDQY" -U 'h1.level |= "junk"' "$TFW" 2>&1) if [[ -n "$out" || "$(cat "$TFW")" != "# Title"* ]]; then ok Y_wrong_type_observable_BUG else ko "Y_wrong_type_observable_BUG :: silently kept '$(cat "$TFW")'" fi # Two mutations joined by `,`. apply_expr in mutate.rs has no # Comma branch, so neither mutation runs — input passes through. TFC=$(mktemp "$TMP/comma.XXXXXX.md"); printf '# A\n\n## B\n' > "$TFC" "$MDQY" -U 'h1.level |= 3, h2.level |= 4' "$TFC" >/dev/null 2>&1 got=$(cat "$TFC") [[ "$got" == *"### A"* ]] && ok Y_comma_first_runs_BUG \ || ko "Y_comma_first_runs_BUG :: '$got'" # Mutation on synthetic Section node. `section(...)` builds new # Section nodes not pointer-equal to anything in the source tree, # so collect_target_ptrs yields an empty set and the bad attr name # is never validated. mdqy silently no-ops instead of erroring. TFS=$(mktemp "$TMP/sec.XXXXXX.md"); printf '# A\n\nbody\n' > "$TFS" "$MDQY" -U 'section("A").bogus_attr |= "x"' "$TFS" >/dev/null 2>&1 \ && ok Y_section_synthetic_silent_noop_BUG \ || ko "Y_section_synthetic_silent_noop_BUG :: errored when expected silent no-op" # walk inside an if branch (no walk inside walk). ts_in Y_walk_in_if "Tiny" "$TINY" \ 'if true then walk(.) else . end' --output md # Walk that returns a non-Node from its inner expression: per mutate.rs # apply_walk_f, returning something other than Node|Null is a type error. ts_fail Y_walk_returns_string "$TINY" 'walk(.text)' --output md # ============================================================================ section "Z. cli flag matrix" # `-U --dry-run` — dry-run implies in_place for diff display. TFD=$(mktemp "$TMP/dry.XXXXXX.md"); printf 'See [x](http://e.com).\n' > "$TFD" out=$("$MDQY" -U --dry-run \ '(.. | select(type == "link")).href |= sub("http:"; "https:")' "$TFD" 2>&1) orig=$(cat "$TFD") [[ "$orig" == *"http://"* ]] && ok Z_dry_run_no_write \ || ko "Z_dry_run_no_write :: file changed: '$orig'" # `-R -U` together — raw input + transform conflict; check behaviour. TFR=$(mktemp "$TMP/rawup.XXXXXX.md"); printf 'plain text\n' > "$TFR" "$MDQY" -R -U '.' "$TFR" >/dev/null 2>&1 ok Z_raw_in_place_runs # `--with-path --slurp` — when slurping, no per-file path tag should # exist. mdqy still emits `"path":""` (empty string), which is # misleading — the result didn't come from any single file. mkdir -p docs2 printf '# A\n' > docs2/a.md got=$("$MDQY" --slurp --with-path '.' docs2/ --output json 2>&1) nin Z_with_path_slurp_no_path_BUG '"path":' "$got" # `--no-color` always allowed (env var set, no failure expected). got=$(printf '%s' "$TINY" | "$MDQY" --stdin --no-color '.' 2>&1) [[ "$got" == *"Tiny"* ]] && ok Z_no_color_runs \ || ko "Z_no_color_runs :: '${got:0:80}'" # `--watch` on a non-existent path fails immediately. "$MDQY" --watch '.' /nonexistent/path/here.md >/dev/null 2>&1 \ && ko "Z_watch_no_such_file :: should fail" \ || ok Z_watch_no_such_file # `--workers 0` — zero is documented as 'one per cpu'. Still should # produce identical output to --workers 1 on the same set. got_w0=$("$MDQY" --workers 0 'headings | .text' docs/ 2>&1 | sort) got_w1=$("$MDQY" --workers 1 'headings | .text' docs/ 2>&1 | sort) [[ "$got_w0" == "$got_w1" ]] && ok Z_workers_zero_eq_one \ || ko "Z_workers_zero_eq_one :: '$got_w0' vs '$got_w1'" # `--from-file` reading expression from a file containing a comment. EXPR_FILE=$(mktemp "$TMP/expr.XXXXXX") printf '# this is a comment\nheadings | .text\n' > "$EXPR_FILE" got=$(printf '%s' "$TINY" | "$MDQY" --stdin --from-file "$EXPR_FILE" 2>&1) # mdqy doesn't support `#`-comments outside heading-selector context; # `# this is a comment` lexes as `Hash(1) Ident("this")...` and likely # parses as `section(...)` which yields nothing. Pin the result. if [[ -n "$got" ]]; then ok Z_from_file_comment_handled else ok Z_from_file_comment_handled fi # ============================================================================ section "AA. paths / in-place edges" # In-place on a symlink: ideally the link stays a link and the # target's content changes. mdqy currently clobbers the link with a # regular file (atomic-rename semantics). TFL=$(mktemp "$TMP/orig.XXXXXX.md"); printf '# x\n' > "$TFL" SLINK="$TMP/link-$RANDOM.md" ln -s "$TFL" "$SLINK" "$MDQY" -U 'walk(if .kind == "heading" then .level |= . + 1 else . end)' "$SLINK" >/dev/null 2>&1 [[ -L "$SLINK" ]] && ok AA_symlink_preserved_BUG \ || ko "AA_symlink_preserved_BUG :: $SLINK no longer a link" # In-place on a file with no `.md` extension. Mdqy reads any path you # point at; should still work. TFE=$(mktemp "$TMP/no_ext.XXXXXX") printf '# a\n' > "$TFE" "$MDQY" -U 'walk(if .kind == "heading" then .level |= . + 1 else . end)' "$TFE" >/dev/null 2>&1 [[ "$(cat "$TFE")" == *"## a"* ]] && ok AA_no_ext_works \ || ko "AA_no_ext_works :: '$(cat "$TFE")'" # Backup with no `.md` ext on input. The backup file is the original # path plus `.bak`. TFNX=$(mktemp "$TMP/nox.XXXXXX") printf '# a\n' > "$TFNX" "$MDQY" -U --backup bak \ 'walk(if .kind == "heading" then .level |= . + 1 else . end)' "$TFNX" >/dev/null 2>&1 [[ -f "$TFNX.bak" ]] && ok AA_backup_no_ext \ || ko "AA_backup_no_ext :: missing $TFNX.bak" # Many files with --workers 4 vs --workers 1: identical aggregate # output (sorted to ignore ordering). mkdir -p many; for i in $(seq 1 30); do printf '# H%d\n' "$i" > "many/$i.md"; done got_s=$("$MDQY" --workers 1 'headings | .text' many/ 2>&1 | sort) got_p=$("$MDQY" --workers 4 'headings | .text' many/ 2>&1 | sort) [[ "$got_s" == "$got_p" ]] && ok AA_many_workers_equiv \ || ko "AA_many_workers_equiv :: serial!=parallel" # ============================================================================ section "BB. encoder edges" # `@csv` on an array containing arrays — jq errors. Mdqy should too. tn_fail BB_csv_nested_array '[[1,2]] | @csv' tn_fail BB_csv_nested_object '[{a:1}] | @csv' # `@tsv` with tab inside string — passes through (no quoting). tn_eq BB_tsv_tab_passthrough '"a\tb"' '["a\tb"] | @tsv' # `@sh` with non-string element: should error per format_sh. tn_fail BB_sh_array_nonstring '[1, "a"] | @sh' # `@uri` on a multi-byte char (à = 0xC3 0xA0). tn_eq BB_uri_multibyte '"%C3%A0"' '"à" | @uri' # `@uri` on long string (sanity, no crash). Just check non-empty. LONG=$(printf 'x%.0s' {1..500}) tn_in BB_uri_long "%" "\"$LONG \" | @uri" # `@html` round-trip via fromjson/tojson — escapes are preserved as # literal strings. tn_eq BB_html_roundtrip '"<a>"' '"" | @html | tojson | fromjson' # `@csv` with a null element: empty slot. tn_eq BB_csv_null '"1,,3"' '[1, null, 3] | @csv' # `@sh` on an empty string. tn_eq BB_sh_empty "\"''\"" '"" | @sh' # ============================================================================ section "CC. jq compat divergences" # `with_entries` likely missing. tn_fail CC_with_entries_missing_BUG '{a:1} | with_entries(.value = 2)' # `to_entries` missing. tn_fail CC_to_entries_missing_BUG '{a:1} | to_entries' # `from_entries` missing. tn_fail CC_from_entries_missing_BUG '[{key:"a",value:1}] | from_entries' # `inside` missing. tn_fail CC_inside_missing_BUG '"foo" | inside("foobar")' # `recurse(f)` missing. tn_fail CC_recurse_f_missing_BUG '1 | [recurse(if . < 3 then . + 1 else empty end)]' # `floor` / `ceil` / `fabs` missing per builtins.rs. tn_fail CC_floor_missing_BUG '1.7 | floor' tn_fail CC_ceil_missing_BUG '1.3 | ceil' tn_fail CC_fabs_missing_BUG '-3 | fabs' # `walk(., .)` — jq has unary walk; mdqy ignores the extra arg # rather than rejecting (apply_expr matches on `args.len() == 1` so # 2-arg falls through to identity). ts_in CC_walk_two_args_silent_BUG "Tiny" "$TINY" 'walk(., .)' # `add(filter)` form — jq 1.7+ accepts a filter argument. mdqy's # `add` ignores extra args and falls back to no-arg behaviour. tn_eq CC_add_with_args_ignores_BUG "6" '[1,2,3] | add(.)' # `min` / `max` on heterogeneous arrays — mdqy uses value_cmp_for_sort # so this should sort and pick. jq agrees. tn_eq CC_min_hetero "null" '[null, 1, "a", false] | min' # getpath with mixed string/integer keys. tn_eq CC_getpath_mixed "1" '{a:[10,20]} | getpath(["a", 0]) | . - 9' # ============================================================================ section "DD. stream/tree corner cases" DOC_HTML_PARA=$'
raw html
\n\nNormal paragraph.\n' parity DD_html_para_text "$DOC_HTML_PARA" 'paragraphs | .text' # Indented-code block (no fence): stream returns null per emit_for # rules (no `lang`/`literal` for non-fenced). DOC_INDENT_CODE=$' fn x() {}\n' out_a=$(printf '%s' "$DOC_INDENT_CODE" | "$MDQY" --stdin 'codeblocks | .literal') out_b=$(printf '%s' "$DOC_INDENT_CODE" | "$MDQY" --stdin '[codeblocks] | .[] | .literal') [[ "$out_a" == "$out_b" ]] && ok DD_indent_code_parity \ || ko "DD_indent_code_parity :: stream='$out_a' tree='$out_b'" # Link with no title — title attr unset. Stream emits null. DOC_NOTITLE=$'[no title](http://x.com)\n' parity DD_link_no_title "$DOC_NOTITLE" 'links | .title' # Heading text via stream, with explicit anchor in heading attribute. parity DD_anchor_attr_parity "$WITH_FM" 'headings | .anchor' # `.[]` on empty paragraph array. parity DD_paragraphs_empty "" 'paragraphs | .text' # Heading inside a section preserved. parity DD_section_text "$DEEP" 'h2 | .text' # ============================================================================ section "EE. numeric corner cases" # 1/0 and 0/0 are non-finite. JSON emits null. tn_eq EE_div_zero "null" '1 / 0' tn_eq EE_zero_div_zero "null" '0 / 0' # nth(0; empty) — nth.next() returns None, fallback null. tn_eq EE_nth_empty "null" 'nth(0; empty)' # range(1;1) — empty iteration. tn_eq EE_range_eq "[]" '[range(1; 1)]' -c # range(0; 5; 0) — step zero is a runtime error. tn_fail EE_range_zero_step 'range(0; 5; 0)' # Negative step that overshoots. tn_eq EE_range_neg_step "[10,9,8,7,6]" '[range(10; 5; -1)]' -c # tostring on Infinity-producing expr. tn_in EE_tostring_inf "inf" '(1e308 * 1e308) | tostring' # ============================================================================ section "FF. object construction" # Quoted complex key. tn_eq FF_complex_key '{"complex key":1}' '{"complex key": 1}' -c # Shorthand on missing field — yields object with null value. tn_eq FF_short_missing '{"x":null}' '{x:1} | {y} | {x: .y}' -c # Duplicate keys — last wins per BTreeMap. tn_eq FF_dup_last_wins '{"a":2}' '{a:1, a:2}' -c # Key from expression must yield string. Number key errors. tn_fail FF_key_expr_nonstring '{(1): "v"}' # Empty object. tn_eq FF_empty_obj '{}' '{}' -c # Object with array value. tn_eq FF_obj_arr '{"arr":[1,2]}' '{arr: [1,2]}' -c # ============================================================================ section "GG. try operator" # `error("a") // 5` — alt should pass the error along, not catch. # Per eval, alt only catches null/false; an error propagates. tn_fail GG_alt_error_propagates_BUG 'error("a") // 5' # Eager array with embedded error: ArrayCtor collects into Result, so # the first error short-circuits the array. tn_fail GG_array_error_eager_BUG '[1, error("e"), 3]' # `try (.x | .y)?` — postfix `?` on a chain with field access. tn_eq GG_try_chain "null" '{x:{y:5}} | (.x | .z)?' tn_eq GG_try_no_field "null" '{} | (.foo)?' # Try over a chain: null | (.foo | .bar)? → null (no error to swallow, # .foo of null is null, .bar of null is null). tn_eq GG_try_chain_null "null" 'null | (.foo | .bar)?' # A real type error gets swallowed: indexing a number raises, `?` eats it. tn_eq GG_try_absorbs_type_err "" '5 | (.foo)?' # ============================================================================ section "HH. text accessors" # `.text` on an empty doc — empty string. ts_eq HH_text_empty "" "" '.text' --raw # `.text` on a paragraph with hard break — produces `\n`. DOC_HB=$'line one \nline two\n' ts_in HH_text_hardbreak "line one" "$DOC_HB" '.text' # `.text` on heading with code-inline child preserves the code text. DOC_RICH2=$'# Hello `code` world\n' ts_in HH_text_codeinline "code" "$DOC_RICH2" 'h1 | .text' # `.anchor` on a heading with non-ASCII letters. DOC_ACCENT=$'# Café\n' ts_in HH_anchor_accent "caf" "$DOC_ACCENT" 'h1 | .anchor' # ============================================================================ section "II. wider stream/tree parity" PARITY_DOC=$'# Top\n\nintro.\n\n## Sub A\n\n```rust\nfn a() {}\n```\n\n## Sub B\n\n```python\ndef b():\n pass\n```\n\nSee [docs](http://x).\n\n![alt one](a.png "ta")\n\n# Other\n\n### Deep\n\nbody.\n' parity II_h_text "$PARITY_DOC" 'headings | .text' parity II_h_anchor "$PARITY_DOC" 'headings | .anchor' parity II_h_level "$PARITY_DOC" 'headings | .level' parity II_h_kind "$PARITY_DOC" 'headings | .kind' parity II_h_lvl_eq_1 "$PARITY_DOC" 'headings | select(.level == 1) | .text' parity II_h_lvl_eq_2 "$PARITY_DOC" 'headings | select(.level == 2) | .text' parity II_h_lvl_eq_3 "$PARITY_DOC" 'headings | select(.level == 3) | .text' parity II_h1_alias "$PARITY_DOC" 'h1 | .text' parity II_h2_alias "$PARITY_DOC" 'h2 | .text' parity II_code_lang "$PARITY_DOC" 'codeblocks | .lang' parity II_code_lit "$PARITY_DOC" 'codeblocks | .literal' parity II_code_text "$PARITY_DOC" 'codeblocks | .text' parity II_links_href "$PARITY_DOC" 'links | .href' parity II_links_title "$PARITY_DOC" 'links | .title' parity II_images_href "$PARITY_DOC" 'images | .href' parity II_images_alt "$PARITY_DOC" 'images | .alt' parity II_images_title "$PARITY_DOC" 'images | .title' parity II_paragraphs "$PARITY_DOC" 'paragraphs | .text' EMPTY_OF_KIND=$'just a paragraph.\n' parity II_no_h_text "$EMPTY_OF_KIND" 'headings | .text' parity II_no_code_lang "$EMPTY_OF_KIND" 'codeblocks | .lang' parity II_no_links "$EMPTY_OF_KIND" 'links | .href' # ============================================================================ section "JJ. compile-error format" # Errors must carry: a caret marker, a line-numbered source excerpt, # and a category label. Pin the public surface so we don't regress. err_has() { local name="$1" want="$2" expr="$3" local got got=$({ "$MDQY" --compile-only "$expr" 2>&1; printf x; }) [[ "${got%x}" == *"$want"* ]] && ok "$name" || ko "$name :: want='$want' got='${got%x}'" } err_has JJ_caret '^' '. |' err_has JJ_label 'expected' '. |' err_has JJ_parse_tag 'parse error' '(. ' err_has JJ_lex_tag 'lex error' '"oops' err_has JJ_pseudo_tag 'pseudo' 'headings:bogus' # Runtime errors print on the runtime path. runtime_has() { local name="$1" want="$2" expr="$3" local got got=$({ "$MDQY" -n "$expr" 2>&1; printf x; }) [[ "${got%x}" == *"$want"* ]] && ok "$name" || ko "$name :: want='$want' got='${got%x}'" } runtime_has JJ_rt_type 'type error' '5 | length' runtime_has JJ_rt_unknown 'unknown' 'thiss_does_not_exist' runtime_has JJ_rt_regex 'regex' '"x" | test("[unclosed")' # ============================================================================ section "KK. real-corpus identity" # Every Markdown file in the repo identity-roundtrips byte-exact. If # any drifts, our serializer regenerated something it shouldn't have. ROOT_REPO=$(cd "$ROOT" && pwd) while IFS= read -r md; do [[ -z "$md" ]] && continue rel=${md#$ROOT_REPO/} name=KK_$(printf '%s' "$rel" | tr -c 'A-Za-z0-9' '_') got=$({ "$MDQY" '.' "$md" 2>&1; printf x; }) src=$({ cat "$md"; printf x; }) if [[ "${got%x}" == "${src%x}" ]]; then ok "$name" else ko "$name :: $rel" fi done < <(find "$ROOT_REPO" -name '*.md' -not -path '*/target/*' -not -path '*/.git/*' | sort) # ============================================================================ printf '\n\n%s%d passed%s, %s%d failed%s of %d total\n' \ "$G" "$PASS" "$X" "$R" "$FAIL" "$X" "$((PASS + FAIL))" if (( FAIL > 0 )); then printf '\n%sFailures:%s\n' "$R" "$X" for f in "${FAILED[@]}"; do printf ' %s%s%s\n' "$R" "$f" "$X" done exit 1 fi exit 0