Data & I/O

JSON / CSV / XML

the data-format conversion builtin family (§ format conversion builtins, 2026-09-01, Ahmed's ask -- BACKLOG.md's "Format conversion builtins" entry): three formats (JSON/CSV/XML), each with an `xify(value) -> str` / `parse_x(str) -> value` primitive pair,

catalog/qu_format_conversion.qu

plus pairwise converters that are one-line compositions of those primitives (`json2csv(s) = csvify(parse_json(s))`, etc.) -- not six independent implementations.

What it exercises

jsonify(value) / parse_json(str) -- Value <-> JSON string
csvify(table) / parse_csv(str) -- Table <-> CSV string
xmlify(value) / parse_xml(str) -- Value <-> XML string (the new piece)
csv2json(str), json2xml(str) -- two of the six pairwise converters
Every check below compares an ACTUAL value read back out of the round
trip against the ORIGINAL data, printed side by side -- not just "it ran
without error".

How it works

JSON: a plain number and a vector round-trip exactly

n = 3.14159265358979
n_json = jsonify(n)
n_back = parse_json(n_json)
print("n_json = {n_json}")
print("n round-trips exactly: {n == n_back}")

v = [10, 20, 30.5]
v_json = jsonify(v)
v_back = parse_json(v_json)
print("v_json = {v_json}")
print("v round-trips exactly: {v[0] == v_back[0] and v[1] == v_back[1] and v[2] == v_back[2]}")

CSV: real sample data, hand-written, parsed into a Table and back

csv_text = "name,age,city
Alice,30,Paris
Bob,25,Tunis
"
t = parse_csv(csv_text)
print("parsed table: {nrow(t)} rows x {ncol(t)} cols, columns = {t.name}, {t.age}, {t.city}")
print("row 0: name={t.name[0]}, age={t.age[0]}, city={t.city[0]}")
print("row 1: name={t.name[1]}, age={t.age[1]}, city={t.city[1]}")

back_csv = csvify(t)
t2 = parse_csv(back_csv)
same_shape = (nrow(t) == nrow(t2)) and (ncol(t) == ncol(t2))
same_data = (t.name[0] == t2.name[0]) and (t.age[0] == t2.age[0]) and (t.name[1] == t2.name[1]) and (t.age[1] == t2.age[1])
print("csvify -> parse_csv round trip preserves shape and data: {same_shape and same_data}")

XML: the same table, converted to XML text and back -- print the raw

XML so the row-oriented shape (one <row> per record, one child element per column, a per-column `kind` attribute) is visible, not just asserted.

xml_text = xmlify(t)
print("--- xmlify(t) output ---")
print(xml_text)
print("--- end xmlify(t) output ---")

t3 = parse_xml(xml_text)
xml_round_trip_ok = (nrow(t3) == nrow(t)) and (t3.name[0] == t.name[0]) and (t3.age[1] == t.age[1]) and (t3.city[1] == t.city[1])
print("xmlify -> parse_xml round trip preserves data: {xml_round_trip_ok}")

# A plain record round-trips through XML too (Num + Str fields, wrapped one
# level down under their own field-name element -- see xml_ops.rs's own
# module doc comment for the full Value<->XML mapping).
r = {city = "Sfax", population = 955421}
r_xml = xmlify(r)
r_back = parse_xml(r_xml)
print("record xmlify output: {r_xml}")
print("record round-trips exactly: {r.city == r_back.city and r.population == r_back.population}")

Pairwise converters -- one-line compositions, exercised on real data

csv_for_pairwise = "product,price
Widget,9.99
Gadget,19.5
"
j_from_csv = csv2json(csv_for_pairwise)
print("csv2json output: {j_from_csv}")
back_via_json = parse_json(j_from_csv)
print("csv2json round trip: product[0]={back_via_json.product[0]}, price[1]={back_via_json.price[1]}")

t_json = jsonify(t)
xml_from_json = json2xml(t_json)
back_via_xml = parse_xml(xml_from_json)
json2xml_ok = (nrow(back_via_xml) == nrow(t)) and (back_via_xml.name[0] == t.name[0]) and (back_via_xml.age[0] == t.age[0])
print("json2xml(jsonify(t)) -> parse_xml recovers the same table: {json2xml_ok}")

A genuinely-can't-convert case: a bare scalar has no CSV/tabular shape

try
    csvify(42)
catch err
    print("csvify(42) correctly refuses with: {err.message}")
end

# ...and a live handle can't become JSON/XML data, either -- the same
# clear refusal `save`/`save_all` already give (jsonify/xmlify reuse it
# verbatim rather than a bespoke message).
function noop()
    return 0
end function
w = spawn("noop")
try
    jsonify(w)
catch err
    print("jsonify(a worker handle) correctly refuses with: {err.message}")
end
join(w)

Results

printed

n_json = {"type":"num","v":3.14159265358979}
n round-trips exactly: true
v_json = {"type":"vec","v":[10.0,20.0,30.5]}
v round-trips exactly: true
parsed table: 2 rows x 3 cols, columns = ("Alice", "Bob"), [30, 25], ("Paris", "Tunis")
row 0: name=Alice, age=30, city=Paris
row 1: name=Bob, age=25, city=Tunis
csvify -> parse_csv round trip preserves shape and data: true
--- xmlify(t) output ---
<table>
  <columns>
    <column kind="str">name</column>
    <column kind="num">age</column>
    <column kind="str">city</column>
  </columns>
  <row>
    <name>Alice</name>
    <age>30</age>
    <city>Paris</city>
  </row>
  <row>
    <name>Bob</name>
    <age>25</age>
    <city>Tunis</city>
  </row>
</table>
--- end xmlify(t) output ---
xmlify -> parse_xml round trip preserves data: true
record xmlify output: <record>
  <city>
    ... 12 more lines

The whole program

Runnable as it stands: qu run catalog/qu_format_conversion.qu

# qu_format_conversion.qu -- the data-format conversion builtin family

# (§ format conversion builtins, 2026-09-01, Ahmed's ask -- BACKLOG.md's

# "Format conversion builtins" entry): three formats (JSON/CSV/XML), each

# with an `xify(value) -> str` / `parse_x(str) -> value` primitive pair,

# plus pairwise converters that are one-line compositions of those

# primitives (`json2csv(s) = csvify(parse_json(s))`, etc.) -- not six

# independent implementations.

#

# Constructs exercised (all real, verified-working builtins):

#   jsonify(value) / parse_json(str)   -- Value <-> JSON string

#   csvify(table) / parse_csv(str)     -- Table <-> CSV string

#   xmlify(value) / parse_xml(str)     -- Value <-> XML string (the new piece)

#   csv2json(str), json2xml(str)       -- two of the six pairwise converters

#

# Every check below compares an ACTUAL value read back out of the round

# trip against the ORIGINAL data, printed side by side -- not just "it ran

# without error".


backend auto

#%% JSON: a plain number and a vector round-trip exactly

n = 3.14159265358979
n_json = jsonify(n)
n_back = parse_json(n_json)
print("n_json = {n_json}")
print("n round-trips exactly: {n == n_back}")

v = [10, 20, 30.5]
v_json = jsonify(v)
v_back = parse_json(v_json)
print("v_json = {v_json}")
print("v round-trips exactly: {v[0] == v_back[0] and v[1] == v_back[1] and v[2] == v_back[2]}")

#%% CSV: real sample data, hand-written, parsed into a Table and back

csv_text = "name,age,city
Alice,30,Paris
Bob,25,Tunis
"
t = parse_csv(csv_text)
print("parsed table: {nrow(t)} rows x {ncol(t)} cols, columns = {t.name}, {t.age}, {t.city}")
print("row 0: name={t.name[0]}, age={t.age[0]}, city={t.city[0]}")
print("row 1: name={t.name[1]}, age={t.age[1]}, city={t.city[1]}")

back_csv = csvify(t)
t2 = parse_csv(back_csv)
same_shape = (nrow(t) == nrow(t2)) and (ncol(t) == ncol(t2))
same_data = (t.name[0] == t2.name[0]) and (t.age[0] == t2.age[0]) and (t.name[1] == t2.name[1]) and (t.age[1] == t2.age[1])
print("csvify -> parse_csv round trip preserves shape and data: {same_shape and same_data}")

#%% XML: the same table, converted to XML text and back -- print the raw

# XML so the row-oriented shape (one <row> per record, one child element

# per column, a per-column `kind` attribute) is visible, not just asserted.

xml_text = xmlify(t)
print("--- xmlify(t) output ---")
print(xml_text)
print("--- end xmlify(t) output ---")

t3 = parse_xml(xml_text)
xml_round_trip_ok = (nrow(t3) == nrow(t)) and (t3.name[0] == t.name[0]) and (t3.age[1] == t.age[1]) and (t3.city[1] == t.city[1])
print("xmlify -> parse_xml round trip preserves data: {xml_round_trip_ok}")

# A plain record round-trips through XML too (Num + Str fields, wrapped one

# level down under their own field-name element -- see xml_ops.rs's own

# module doc comment for the full Value<->XML mapping).

r = {city = "Sfax", population = 955421}
r_xml = xmlify(r)
r_back = parse_xml(r_xml)
print("record xmlify output: {r_xml}")
print("record round-trips exactly: {r.city == r_back.city and r.population == r_back.population}")

#%% Pairwise converters -- one-line compositions, exercised on real data

csv_for_pairwise = "product,price
Widget,9.99
Gadget,19.5
"
j_from_csv = csv2json(csv_for_pairwise)
print("csv2json output: {j_from_csv}")
back_via_json = parse_json(j_from_csv)
print("csv2json round trip: product[0]={back_via_json.product[0]}, price[1]={back_via_json.price[1]}")

t_json = jsonify(t)
xml_from_json = json2xml(t_json)
back_via_xml = parse_xml(xml_from_json)
json2xml_ok = (nrow(back_via_xml) == nrow(t)) and (back_via_xml.name[0] == t.name[0]) and (back_via_xml.age[0] == t.age[0])
print("json2xml(jsonify(t)) -> parse_xml recovers the same table: {json2xml_ok}")

#%% A genuinely-can't-convert case: a bare scalar has no CSV/tabular shape

try
    csvify(42)
catch err
    print("csvify(42) correctly refuses with: {err.message}")
end

# ...and a live handle can't become JSON/XML data, either -- the same

# clear refusal `save`/`save_all` already give (jsonify/xmlify reuse it

# verbatim rather than a bespoke message).

function noop()
    return 0
end function
w = spawn("noop")
try
    jsonify(w)
catch err
    print("jsonify(a worker handle) correctly refuses with: {err.message}")
end
join(w)