JSON Automation


maXbox Starter 85 — JSON Automation with Json4Delphi

There are two kinds of data scientists:

1) Those who can extrapolate from incomplete data.

Reading JSON data in maXbox or Lazarus should be easy with the right class. JSON can be read from a string, file or it could be a JSON link see later on.

But what’s JSON: JSON (JavaScript Object Notation) is a lightweight data-interchange format. It is easy for humans to read and write at least as a text. It is easy for machines to parse and generate, but not so easy to interpret for humans.

Lets start with a simple sample:

First we create an object and parse it:

ajt:= TJson.create();


//Now in Json4Delphi we can ask the type:

writeln(botostr( ajt.IsJsonObject( stjson)));

writeln(botostr( ajt.IsJsonString( stjson)));

writeln(botostr( ajt.IsJsonArray( stjson)));

cnode:= ajt.JsonObject.items[0].name;


As you can see the sample is an object node and data is the cnode.

Creta Base Jump

JSON for Delphi supports also older versions of Delphi (7 or above) and its Object Pascal native code, using classes like TList, TStrings and TStringList is a great advantage for speed, scripting and comprehension.

So how do we get the branch in our example:

writeln(‘branch of data: ‘+ajt[‘data’].asobject[‘results’].asarray[0].asObject[‘Branch’].asstring);

So the branch is an object-array. Arrays in JSON are almost the same as arrays in Pascal or C. In JSON, array values must be of type string, number, object, array, boolean or null. In JavaScript, array values can be all of the above, plus any other valid JavaScript expression, including functions, dates or undefined. In Delphi we use of course strong types with overloading functions not dynamic string types!


TJsonValueType = (jvNone, jvNull, jvString, jvNumber, jvBoolean,
jvObject, jvArray);

TJsonStructType = (jsNone, jsArray, jsObject);

TJsonNull = (null);

TJsonEmpty = (empty);

On the other side JSON is a text format for representing objects and arrays, there is no such thing as a “JSON object” like a Object Pascal Object. Therefore we have to find out in our J4D library the type from the formal syntax:

function TJsonBaseAnalyzeJsonValueType(const S: String): TJsonValueType;

var Len: Integer; Number: Extended;


Result:= jvNone;

Len:= Length(S);

if Len >= 2 then begin

if (S[1] = ‘{‘) and (S[Len] = ‘}’) then Result := jvObject

else if (S[1] = ‘[‘) and (S[Len] = ‘]’) then Result := jvArray

else if (S[1] = ‘“‘) and (S[Len] = ‘“‘) then Result := jvString

else if SameText(S, ‘null’) then Result := jvNull

else if SameText(S, ‘true’) or SameText(S, ‘false’) then Result:= jvBoolean

else if FixedTryStrToFloat(S, Number) then Result := jvNumber;


else if FixedTryStrToFloat(S, Number) then Result := jvNumber;


Next topic is a Json-tree. Normally the packed collection data we use is imported from a file or folder but we can also parse and stringify a const as json4delphi data or test data:

Again we can see the formal syntax. Similar to other formed programming languages, an Array in JSON is a list of items surrounded in square brackets ([]). Each item in the array is separated by a comma.

A JSON object (a string to parse you remember) is a key-value data format that is typically rendered in curly braces{}. Our JSON object above looks something like this:

JSON arrays are ordered collections and can contain values of different data types and this is more flexible than in XML. I don’t think that JSON syntax is very complicated and I prefer it over XML and YAML.

Ok. lets do 2 ways of accessing our distance map data from above:

ajt:= TJson.create();


writeln(botostr( ajt.IsJsonObject(StrJson)));

writeln(botostr( ajt.IsJsonString(StrJson)));

writeln(botostr( ajt.IsJsonArray(StrJson)));

writeln(‘get third name: ‘+ ajt.JsonObject.items[2].name);

writeln(‘get four name: ‘+ ajt.JsonObject.items[3].name);

println(‘dist: ‘+ajt[‘rows’].asarray[0].asObject[‘elements’].asarray[0].asobject[‘distance’].asobject[‘text’].asstring);

We can also access array or multi-dimensional array values by using a for loop and index numbers:

jOb:= ajt.JsonObject; //reference passing

for cnt:= 2 to jOb.count-2 do begin

Clabel:= job.items[cnt].name;

writeln(‘iterate: ‘+clabel)

JsArr:= job.values[Clabel].asArray;

for cnt2:= 0 to jsarr.count-1 do

jsobj:= jsarr.items[cnt2].asobject;

for cnt3:= 0 to jsobj.count do




If you prefer direct access for example of the status:

println(‘elements status: ‘+ajt[‘rows’].asarray[0].asObject[‘elements’].asarray[0].asobject[‘status’].asstring);

For a big data collection its important to know your memory allocation and free them as many as possible or keep the object lifetime short:

memory leaks
Dataleak of Jsonvalues

pic: m85_1026_dataleakmark.png

As a next and last sad example we get the data from web.

Let us first try to read the JSON data from a web link.

JsonUrl = 'https://pomber.github.io/covid19/timeseries.json';

Now we need a Load URL() or Upload File() function to get the JSON data for parsing. In our case load is a ole automation function-pair of open and send(). We define the necessary packages “msxml2.xmlhttp” and the JSON class itself:

var XMLhttp : OleVariant; // As Object

ajt: TJson; JObj: TJsonObject2;

XMLhttp:= CreateOleObject(‘msxml2.xmlhttp’)

XMLhttp.Open (‘GET’, JsonUrl, False)

ajt:= TJson.create();

Let us import the covid19 timeseries data from this already mentioned JSON link: pomber.github.io/covid19/timeseries.json using XMLhttp:

Ref:  <class 'pandas.core.frame.DataFrame'>
RangeIndex: 82661 entries, 0 to 82660
Data columns (total 5 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 country 82661 non-null object
1 date 82661 non-null object
2 confirmed 82661 non-null int64
3 deaths 82661 non-null int64
4 recovered 82661 non-null int64
dtypes: int64(3), object(2)
memory usage: 3.4+ MB

A JSON Parser is then used to format the JSON data into a properly and readable JSON Format with curly brackets. That can easily view and identify its key and values. To get the JSON type of class, struct or array, we need to use ajt.parse() method first.

For slicing (filter) the data we copy the range from response timeseries.json:

start:= pos(‘“‘+ACOUNTRY+’”’,response);

stop:= pos(‘“‘+ACOUNTRY2+’”’,response);

writeln(‘Len Overall: ‘+itoa(length(response)))

resrange:= Copy(response, start, stop-start);

resrange:= ‘{‘+resrange+’}’;

writeln(‘debug sign on pos: ‘+GetWordOnPos(response, posex(‘],’,response,1)));




writeln( ‘Exception: <TJson>”” parse error: {‘+

exceptiontostring(exceptiontype, exceptionparam))



writeln(‘Statuscode: ‘+(statuscode)+’: ‘+’listlen ‘+itoa(slist.count));

Now we can iterate through the keys with values as items. Here, in the above sample JSON data: date, confirmed, deaths and recovered are known as key and “2020–1–22”, 0, 0 and 0 known as a Value. All Data are available in a Key and value pair.

First we get a list of all 192 country names as the node name:

JObj:= ajt.JsonObject;

writeln(‘Get all Countries: ‘)

for cnt:= 0 to jobj.count-1 do writeln(Jobj.items[cnt].name);
...United Kingdom

So the country is an object to get. Ok, it is a Json-Object dictionary with 192 countries. We check the keys of our dict with a nested loop of all confirmed cases:

for cnt:= 0 to Jobj.count-1 do begin 
Clabel:= Jobj.items[cnt].name;
JArray2:= jobj.values[Clabel].asArray;
for cnt2:= 0 to jarray2.count-1 do
itmp:= jarray2.items[cnt2].asObject.values['confirmed'].asinteger;

//*) //accumulated
Chart1.Title.Text.add(‘Sciplot TimeSerie for: ‘+Clabel);

JArray:= ajt.values[Clabel].asarray;

writeln(‘jitems country ‘+itoa(jarray.count));

for cnt:= 1 to jarray.count-1 do begin

itmp:= jarray.items[cnt].asObject.values[‘confirmed’].asinteger;


itmp:= jarray.items[cnt].asObject.values[‘deaths’].asinteger;


itmp:= jarray.items[cnt].asObject.values[‘recovered’].asinteger;



sciplot time series
actual covid situation
actual covid situation
maxbox sciplot 192 countries accumulated deaths
pic: m85_covid5_85.png
3 series overlay

Pic: m85_deathconfratio_json.png


The proper way to use JSON is to specify types that must be compatible at runtime in order for your code to work correctly.

The TJsonBase= class(TObject) and TJsonValue= class(TJsonBase) namespace contains all the entry points and the main types.

The TJson= class(TJsonBase) namespace contains attributes and APIs for advanced scenarios and customization.

JSON is a SUB-TYPE of text but not text alone. Json is a structured text representation of an object (or array of objects). We use JSON for Delphi framework (json4delphi), it supports older versions of Delphi and Lazarus (6 or above) and is very versatile. Another advantage is the Object-pascal native code, using classes only TList, TStrings, TStringStream, TCollection and TStringList; The package contains 3 units: Jsons.pas, JsonsUtilsEx.pas and a project Testunit, available at: https://github.com/rilyu/json4delphi

The script can be found:









Doc: https://maxbox4.wordpress.com


Appendix: import register log from maXbox4 integration

{* — — — —— — — — — — — — — — — — — — — — — — — *)

procedure SIRegister_Jsons(CL: TPSPascalCompiler);


‘(jvNone,jvNull,jvString,jvNumber,jvBoolean,jvObject,jvArray )’);

CL.AddTypeS(‘TJsonStructType’, ‘( jsNone, jsArray, jsObject )’);

CL.AddTypeS(‘TJsonNull’, ‘( jsnull2 )’);

CL.AddTypeS(‘TJsonEmpty’, ‘( jsempty )’);










Max Kleiner's professional environment is in the areas of OOP, UML and coding - among other things as a trainer, developer and consultant.